SECTION 1. Findings and Declarations.
(a) Artificial intelligence is becoming ubiquitous across the Commonwealth. In particular, the systems are used to impact decisions around housing, employment, healthcare, education, social development and interactions, and access to public services. The outcomes and inferences, the process by which a deployed machine learning model or generative AI system produces outputs, are not transparent to the individuals and residents they affect. It is readily apparent these systems shape the livelihoods and socioeconomic development of Massachusetts residents often without their knowledge.
(b) The invisible use of AI systems in consequential decisions around human life, development, and dignity is the primary risk to an individual's cognitive sovereignty and to democratic governance itself. These systems have the capability to assess residents without disclosure, modify an individual's or collective group's experience based on an assessment, and shape what information they receive, how options or decisions are framed, and what a person comes to believe is possible. This is drawn from the data collected about them through the interaction between human and AI systems and characteristics the system inferred from it; all invisibly and without the individual's knowledge.
(c) Article 89 of the Amendments to the Massachusetts Constitution, specifically the Home Rule Amendment approved by voters in 1966, and Bloom v. Worcester, 363 Mass. 136 (1973) give municipalities the autonomy to draft and enforce ordinances so as they do not conflict with State law and do not interfere with the aggregate intent of the legislature, specifically an intent that courts are permitted to infer even from statutes that do not reference or mention municipalities at all.
(d) The General Court determines that municipalities possess the constitutional authority to require transparency around AI systems operating in their communities. This authority exists whether or not a municipality chooses to exercise it and shall not be taken away silently. Any restriction on that authority must be stated by the General Court, expressly in plain language, such that no court reads a statute that never mentions a municipality as taking that authority away and so that the customary and traditional liberties of the people in the conduct of their local government are not infringed by silence or invisibility.
(e) The General Court further determines that when any physical AI infrastructure, such as a data center, operates in a community; the impacts that cross beyond the facility's site boundaries, including the water it consumes, the demands put on the electric grid, and the emissions of its backup generation are impacts on the community itself. A municipality shall have the authority to require the owner or operator of such a facility to make regular and transparent reports of these impacts. Their reporting must be drawn from readily available data, expressed in common, explainable units. These reports must be provided to the municipality and to the residents of the community. This finding does not limit the right of residents to transparency around the outputs and inferences of any AI system or machine learning algorithm that operates from such a facility when used in consequential decisions around their lives, as provided in these findings and this act. The authority vested in the municipality is in addition to, and does not diminish, any existing authority around land use, zoning, and public health. This authority shall not be taken away as provided in these findings.
(f) The General Court further determines that when AI systems are deployed and used, the interactions generate data around the people and communities they operate in. This data is then collected and retained, and fed back into the training and improving of these models (creating a feedback loop where deployment creates new data and the data improves the model). This loop gives the data generated by residents and their communities real economic value to the owner or operator of said system. The residents of the communities, that generate this data have the right to know when their information goes back into this loop, whether it is retained, used to train or improve AI and machine learning systems, or is shared, such that the value taken from a community is visible to the community itself in accessible and common language.
(g) The General Court further determines that when AI systems are offered to governments at nominal cost, through federal and state purchasing vehicles, these systems arrive in the schools, the police departments, sheriffs' offices, and the service offices of a community faster than a community can deliberate around them. The speed at which they are implemented reduces their ability to defend their own cognitive sovereignty and, in essence, is a choice made for them not by them. A municipality shall have the authority to make its own decisions around the adoption and operation of AI systems in its own operations and services, to evaluate and adopt them, to decline and discontinue them, and to set their terms of use. This authority belongs to the self government of the community itself. No State agency, executive order, regulation, or administrative action shall interfere with it. Any restriction on this authority must be stated by the General Court expressly and in plain language.
(h) The General Court further determines that the same speed and the same nominal cost reach the Commonwealth itself as purchaser. When the state acquires an AI system outside the competitive evaluation, the public accountability, and the disclosure of conflicts of interest that the people require of all other public purchasing, the choice is again made for the residents and not by them. The purchasing of AI systems by the Commonwealth must be held to the standards the people have always required of the purchasing done in their name. A system demonstrated under the conditions of its maker's choosing has not been demonstrated under the conditions of the Commonwealth's use, and the difference between those two sets of conditions is borne by the resident.
(i) The General Court further determines that the administration of the transparency rights, certifications, and community understanding established by this act must be entrusted to an entity captured neither by any single branch of government nor by industry, so that the trust the act is built to create is not undone by the body that administers it.
SECTION 2. Purpose.
The purpose of this Act is to:
(a) establish that every resident of the Commonwealth has the right to know when an AI system is used in a consequential decision around their life, at the moment of interaction where the system meets the person and not merely at the moment the system was trained;
(b) protect the authority of municipalities to require, if they choose, transparency around AI systems operating in their communities, an authority that shall not be diminished by this Act and shall not be taken away except as the General Court provides expressly and in plain language;
(c) recognize that the impacts of physical AI infrastructure cross beyond the boundaries of the facility's site and into the community itself, and protect the authority of municipalities to require regular and transparent reporting of those impacts, drawn from readily available data and expressed in common, explainable terms and units;
(d) ensure that residents know when the data generated by their interactions with an AI system is retained, used to train or improve the system, or shared; and that the value extracted from a community is made visible and transparent to the community itself, in accessible and common language;
(e) ensure that when the Commonwealth purchases an AI system, the purchase is held to the same competitive evaluation, the same public accountability, and the same disclosure of conflicts of interest that the people require of all other public purchasing; and to provide that the Commonwealth, as purchaser, shall require of any vendor a plain and truthful accounting of whether the system can disclose its use, its outputs, and its inferences to the people it is used upon; that the Commonwealth shall have the right to audit the inferences of any such system across the linguistic, socioeconomic, and demographic subgroups of its residents; and that the Commonwealth shall choose the system that can be transparent and honest to the people it serves, so that what cannot be disclosed today for want of technological capability is disclosed tomorrow;
(f) build a statewide understanding, grounded in the communities themselves, of how artificial intelligence is experienced by the residents it is used upon, across the cities and towns of the Commonwealth, recognizing that human understanding must be the cornerstone upon which any approach to AI is built;
(g) establish the Massachusetts AI Trust Authority as an independent public entity, captured neither by any single branch of government nor by industry, to administer the voluntary AI Trust Certification, the Innovation Sandbox, the reciprocity frameworks, and the Community Advisory Board established by this Act, and to receive the accountings and conduct the audits provided for in this Act;
(h) make the Commonwealth the place where trustworthy artificial intelligence is built and deployed, by rewarding transparency over invisibility, and by drawing responsible investment through incentive, collaboration, and reciprocity rather than mandate;
(i) demonstrate that transparency at the moment the system meets the person is feasible to the measure of the technological capability these systems already possess, and that the duty to disclose shall grow as technological capability grows; achievable without the compelled disclosure of trade secrets, training data, or the proprietary internal architectures of any model; and that it protects the cognitive sovereignty of the person and, through it, the human condition itself. This Act offers that model to the governance of AI at the state and federal levels, to complement and to strengthen it;
(j) extend transparency to transactions in which an AI system acts on behalf of a resident, so that the right to know does not diminish or fail as these systems evolve from tools, to agents, to collective swarms; and
(k) ensure that no resident of the Commonwealth is left behind as artificial intelligence reshapes the institutions, services, opportunities, and personal growth of daily life.
SECTION 3. The General Laws are hereby amended by inserting after chapter 93L the following chapter: 93M
CHAPTER 93M. THE MASSACHUSETTS DIGITAL COMMONWEALTH
Section 1. Definitions.
As used in this chapter, the following words shall have the following meanings unless the context clearly requires otherwise:
"Artificial intelligence system" or "AI system", any machine-based system that, for a given set of objectives, generates outputs such as predictions, recommendations, decisions, or content that can influence real or virtual environments, including but not limited to machine learning systems, large language models, automated decision-making systems, and AI agents as defined herein.
"AI agent", an AI system configured to take autonomous actions on behalf of a resident, including but not limited to searching, evaluating, negotiating, selecting, transacting, or communicating, without requiring contemporaneous human confirmation for each action. An AI agent includes a system of multiple AI agents acting in coordination, in sequence, or in concert, whether alike or specialized, however composed and however distributed, and the obligations of this chapter shall apply to such a system as a whole.
"Agent provider", the entity that makes an AI agent available to a resident and that the resident has authorized, directly or through use of the agent, to act on the resident's behalf.
"Authority", the Massachusetts AI Trust Authority established under section 9.
"Certified AI system", an AI system holding a valid AI Trust Certification issued by the Authority under this chapter.
"Community Advisory Board" or "Board", the Community Advisory Board established under section 10.
"Consequential decision", a decision, recommendation, or determination for which an AI system is a substantial factor and which has a material legal or similarly significant effect on the provision or denial to a resident of, or the cost or terms of:
(1) employment or an employment opportunity;
(2) housing;
(3) credit or financial services;
(4) insurance;
(5) education enrollment or an educational opportunity;
(6) healthcare services;
(7) public benefits or essential government services;
(8) legal services; or
(9) engagement with the criminal justice system.
A determination shall also constitute a consequential decision, regardless of whether it falls within the categories enumerated above, if the AI system engages in inference-time profiling of the resident in the course of generating its output; provided, however, that where such a determination does not have a material legal or similarly significant effect on the resident, the obligations arising under section 2 shall be limited to the matters described in section 2(a)(1), (2), (4), and (5), and may be satisfied by a clear and conspicuous disclosure presented within the interface through which the resident interacts with the AI system, applicable in standardized form to all interactions under the same deployment.
The term "consequential decision" does not include determinations made by AI systems performing exclusively: (i) narrow procedural tasks; (ii) pattern detection or anomaly flagging where the output is reviewed by a human who has the authority and the practical opportunity to alter or override the outcome before it affects the resident; or (iii) only the following functions: anti-fraud detection not employing facial recognition, cybersecurity or anti-malware operations, spam filtering, spell-checking, grammar correction, calculation, routine data storage or retrieval, or language translation; provided, however, that any such system that engages in inference-time profiling of a resident shall remain subject to the transparency obligations of section 2 notwithstanding this exclusion.
The Authority may, by regulation, add categories of decisions to this definition based on findings reported by the Community Advisory Board under section 10(g)(4), provided that any such addition satisfies the material legal or similarly significant effect standard established by this section. The Authority may not, by regulation, remove any category enumerated in this section.
"Covered facility", physical AI infrastructure with a design critical information technology load of not less than 5 megawatts, together with any colocated or contiguous physical AI infrastructure under common ownership or control the combined design critical information technology load of which is not less than 5 megawatts.
"Delegated transaction", an action taken by an AI agent on behalf of a resident that commits the resident to a financial obligation, alters a contractual relationship, or affects access to a service, without contemporaneous human confirmation of that specific action.
"Deployer", any person, entity, or government agency that deploys or uses an AI system within the commonwealth, or that deploys or uses an AI system, wherever operated, upon the residents of the commonwealth, whether as a developer, operator, or procuring entity.
"Deployment objective", the objective or objectives established for an AI system in a given deployment, including any objective embodied in instructions, configurations, or parameters provided to the system at or before inference, whether established by the deployer, the vendor, or any other party. A deployment objective is stated as purpose and effect, in accessible and common language, and not as the verbatim content of any instruction, configuration, or parameter, and may be stated in a standardized form applicable to all interactions under the same deployment. The objective of a deployment directed at a resident is not a secret from that resident.
"Determination", the output or conclusion an AI system produces at inference with respect to a resident, as distinct from the decision made concerning the resident.
"Enterprise AI deployment", any procurement, contract, or agreement providing AI system access to more than 500 state employees or affecting services delivered to more than 10,000 residents.
"Facility impact report", a written report of the community impacts of a covered facility as provided in section 4, drawn from readily available data and expressed in common, explainable units.
"Inference", the process by which a deployed AI system generates outputs, including predictions, recommendations, assessments, classifications, or content, in response to inputs, including inputs derived from or about a specific individual. Inference occurs at the point of impact: it is the moment at which a model's capabilities are applied to a real person or situation.
"Inference audit", an examination, conducted by or at the direction of the Authority, of the outputs and inferences of an AI system as deployed, to identify disparities in the performance of those inferences across the linguistic, socioeconomic, and demographic subgroups of the residents of the commonwealth, and across the conditions under which the system is used, including variation in the phrasing, language, context, and channel of the interaction, and including the sustained interactions in which the system is used in practice. An audit conducted only under the conditions of the vendor's own evaluation is not an inference audit.
"Inference-time profiling", the process by which an AI system, during the course of generating an output, assesses, classifies, or infers characteristics about an individual, including but not limited to emotional state, cognitive patterns, behavioral tendencies, demographic attributes, vulnerabilities, or susceptibility to influence, for the purpose of modifying the output delivered to that individual.
"Massachusetts Municipal Association" or "MMA", the nonprofit membership association representing the cities and towns of the commonwealth, or its successor organization, acting in its capacity as a named consultation partner under this chapter.
"Material financial relationship", any relationship in which an individual, or the individual's employer, parent company, or subsidiary, holds an ownership stake, investment interest, contractual partnership, or revenue-sharing arrangement with a vendor or prospective vendor, or has received compensation from such vendor within the preceding 36 months.
"Material modification", with respect to a certified AI system, any change in its training data sources, its inference behavior on standardized benchmarks, its deployment scope, or its user population that the Authority by regulation determines to be material. The Authority shall establish the specific thresholds and criteria for material modification by rule and may update such criteria from time to time.
"Physical AI infrastructure", a data center or other physical facility a principal function of which is to house computing, storage, or networking equipment used in the training of AI systems or in the operation of AI systems, including inference.
"Readily available data", data that the owner or operator of a covered facility collects, generates, or receives in the ordinary course of designing, constructing, or operating the facility, or that the owner or operator reports to any federal or state agency, public utility, water or sewer authority, or independent system operator, together with data derivable from the foregoing without material additional instrumentation.
"Substantial factor", with respect to the role of an AI system in a decision, recommendation, or determination, that the AI system both (i) assists in making the decision, recommendation, or determination, and (ii) is capable of altering the outcome. The mere involvement of an AI system in a peripheral, administrative, or ministerial function that is not capable of altering the outcome shall not constitute a substantial factor.
"Transparency", the disclosure of sufficient information about an AI system's purpose, data inputs, logic, capabilities, limitations, and performance to enable meaningful awareness by affected residents and oversight by governing bodies. Disclosure that cannot be understood by the person it is owed to is not transparency.
"Truthful accounting", a written disclosure, attested by an officer or principal of the vendor, stating in accessible and common language whether, and to what extent, an AI system can disclose its use, its outputs, and its inferences to the people it is used upon, including a statement of what the system cannot yet disclose and why. An accounting that omits a known limitation of the system is not truthful.
"Trust Fund", the Massachusetts AI Trust Fund established under section 13.
"Vendor", any person or entity that offers, sells, licenses, leases, or otherwise provides an AI system, or access to an AI system, to the commonwealth or to any deployer within the commonwealth, whether directly or through a reseller, integrator, or purchasing vehicle.
Section 2. Right to Know.
(a) Every resident of the commonwealth has the right to know when an AI system is used to make, inform, or materially influence a determination in a consequential decision affecting one's life. The right begins at inference. It does not begin at training, and no claim about training data, design, or authorship diminishes it. The right belongs to the resident; the duty to disclose and provide transparent understanding of the outcome extends to any affected individual. The right to know includes:
(1) that an AI system was used in the decision;
(2) the purpose and function of the system, and the deployment objective, stated in commonly understandable language;
(3) the categories of data concerning the resident that the system processed at the time of inference;
(4) whether the system assessed, profiled, or inferred characteristics of the individual in the course of generating its output;
(5) whether information provided by the resident, or otherwise processed at inference, is retained, used to train or improve the AI system, or is shared with any third party;
(6) how long information transmitted during inference is stored;
(7) how often the system's performance is reviewed, and whether those reviews catch changes in how the system behaves after its deployment;
(8) how the system behaves under different conditions, including the conditions under which its performance degrades;
(9) how to contest or challenge the decision, with clear and understandable means to reach a human being with sufficient authority to review and audit the decision; and
(10) the identity of the deployer, and whether any third party was used by the deployer in reaching the determination.
(b) This section grants no authority to restrict, delay, condition, or prohibit the development, deployment, or use of any AI system, and none shall be inferred from it. It establishes a right to visibility, not a right to veto.
(c) Disclosure must follow the point of inference. Where inference occurs as part of a human-AI system interaction, the deployer shall provide disclosure on how the outcome was determined, with any relevant context, in common language, at or before the moment of the determination. Where inference occurs without the person present, the disclosure must travel to them; the deployer shall provide direct notice to the affected individual at or before the moment of the decision, and where notice before is impracticable, within a reasonable time, and never more than 30 days after. A posting on a website or in terms of service may support the required disclosure but does not satisfy it. The disclosure must be explainable, and a disclosure the average individual cannot understand does not satisfy this requirement. A disclosure the person had no reasonable occasion to encounter was never made.
(d) This section requires no disclosure of proprietary algorithms or architectures, training data, model weights, or trade secrets, or any other confidential intellectual property. Documentation or auditing of training data is not required. Transparency under this section is defined as disclosure to inform and provide explainable context, not disclosure with sufficient information to replicate or repeat. The obligations of this section attach solely to the application of the AI system, at inference, to the affected individual.
(e) Delegated transactions.
(1) When an AI agent undertakes a delegated transaction on behalf of a resident, the agent provider shall maintain a reviewable record of the transaction and shall make such record available to the resident in plain language within a reasonable time following the transaction. The record shall include:
(A) a description of the action taken;
(B) the alternatives considered by the agent, if any, and the basis for selecting the action taken;
(C) any inference-time profiling performed on the resident or the resident's agent in the course of the transaction, to the extent known by the agent provider;
(D) any financial or contractual consequences of the action for the resident; and
(E) the identity of the agent provider and any third-party AI systems or platforms involved in the transaction, to the extent known by the agent provider.
(2) The agent provider bears primary responsibility for the disclosure obligations of this subsection for the full chain of AI systems and third parties involved in executing a delegated transaction on behalf of a resident, regardless of the number or identity of intermediary systems.
(3) Agent providers shall disclose to residents, prior to the resident's authorization of the agent, any preferred-vendor relationships, affiliate arrangements, or revenue-sharing incentives that may influence the agent's decisions.
(4) The Authority may establish by regulation thresholds below which the record requirement of paragraph (1) does not apply. Such thresholds shall be designed to exclude routine, low-value, de minimis transactions while preserving the record requirement for transactions of material consequence.
(f) Enforcement.
(1) A knowing and willful failure to provide the disclosures required by this section in connection with a consequential decision or delegated transaction shall constitute an unfair or deceptive act or practice under chapter 93A of the General Laws. A resident who demonstrates that a consequential decision was made affecting them, or that a delegated transaction was undertaken on their behalf, without the disclosures required by this section may seek damages under the procedures and remedies established by said chapter 93A. The deprivation of the disclosures required by this section, including the loss of the opportunity to contest the determination and to reach human review as provided in subsection (a)(9), constitutes a distinct injury to the resident, separate from the violation itself.
(2) A resident may seek injunctive or declaratory relief to compel the disclosures required by this section upon a showing that a deployer or agent provider has materially failed to provide such disclosures and that the resident is or will imminently be subject to a consequential decision or delegated transaction by the AI system. A court may grant preliminary injunctive relief under this paragraph upon a finding that the resident is likely to succeed on the merits and that the balance of equities favors disclosure. The court shall not require a showing of knowing and willful conduct as a condition of injunctive or declaratory relief under this paragraph. Relief under this paragraph shall be limited to requiring compliance with the disclosure obligations of this section and shall not extend to restricting the operation of any AI system.
(3) The attorney general may bring an action for injunctive relief, declaratory relief, or civil penalties under chapter 93A, section 4, to enforce the disclosure requirements of this section against any deployer or agent provider. The attorney general may, prior to commencing an action, provide the deployer or agent provider with written notice of the alleged violation and a reasonable opportunity to achieve compliance with the disclosure requirements of this section. In determining whether to bring an action or to provide an opportunity to achieve compliance, the attorney general may consider the scope and frequency of non-compliance, the number of residents affected, and whether the deployer or agent provider has taken good-faith steps to comply or has been responsive to prior inquiries. Nothing in this paragraph limits the authority of the attorney general under chapter 93A or any other provision of law.
(4) Nothing in this subsection creates a new cause of action for damages beyond those available under chapter 93A; it connects the transparency right established by this section to the existing enforcement framework of the commonwealth while ensuring that injunctive relief remains practically available to affected residents and that the attorney general may address systemic non-compliance.
(5) With respect to a deployer that is a government agency not engaged in trade or commerce within the meaning of chapter 93A, the remedies available under this subsection shall be those provided in paragraph (2), together with any other remedy available under law; and the affected resident shall in addition have the right to review of the determination through the means required by subsection (a)(9).
(g) Vendor support obligation. A vendor that provides an AI system used by a deployer in consequential decisions shall supply to the deployer, in standardized form, accurate information within the vendor's knowledge and control reasonably necessary for the deployer to satisfy the disclosure obligations of this section, and shall update such information upon any material modification to the system. A deployer's disclosure made in good-faith and reasonable reliance on information supplied under this subsection satisfies the deployer's obligation with respect to the matters so supplied, and responsibility for any inaccuracy in such information rests with the vendor. Nothing in this subsection requires the disclosure by a vendor of the materials described in subsection (d).
Section 3. Municipal Authority.
(a) Municipalities of the commonwealth shall possess inherent authority, consistent with the Home Rule Amendment and the general laws, to require transparency regarding AI systems deployed or operating within their jurisdictions that affect their residents. This authority includes the authority to:
(1) require deployers to disclose, upon request, the purpose, general function, and categories of data inputs for AI systems used in consequential decisions affecting municipal residents;
(2) require that residents affected by consequential AI decisions within the municipality have access to the information described in section 2; and
(3) participate in voluntary municipal cooperation frameworks for shared transparency infrastructure.
(b) Municipal authority under this section extends to transparency and visibility. Nothing in this section shall be construed to grant municipalities authority to prohibit, restrict, condition, license, or impose operational requirements upon the development, training, deployment, or use of AI systems. Municipalities may require that AI be visible. Municipalities may not require that AI be absent.
(c) No state agency, executive order, regulation, or administrative action shall restrict, diminish, or preempt the authority of municipalities to require transparency regarding AI systems within their jurisdictions as affirmed by this section.
(d) Rule of construction. No provision of the general laws, whether enacted before or after the effective date of this section, shall be construed to restrict, diminish, or preempt the authority of municipalities to require transparency regarding AI systems within their jurisdictions as affirmed by this section, unless such provision specifically and explicitly states that it is intended to restrict, diminish, or preempt municipal AI transparency authority established under this chapter.
(e) Any ambiguity in a provision of the general laws regarding its effect on municipal AI transparency authority shall be resolved in favor of preserving such authority.
(f) A municipality may seek injunctive relief to enforce the transparency requirements of this section through appropriate proceedings in the superior court.
(g) Nothing in this section shall be construed to limit the authority of the General Court to establish minimum standards for AI transparency that municipalities may meet or exceed.
(h) Municipal self-governance. Nothing in this chapter shall be construed to limit the authority of a municipality to decide whether, and on what terms, it procures, adopts, declines, discontinues, or uses AI systems in its own operations and in the services it delivers, and no state agency, executive order, regulation, or administrative action shall compel such adoption or interfere with such decisions. The limitations of subsection (b) concern the regulation by a municipality of persons other than the municipality itself and do not diminish the authority affirmed by this subsection.
Section 4. Physical AI Infrastructure Transparency.
(a) The impacts of a covered facility cross beyond the boundaries of the facility's site and into the community itself. A municipality in which a covered facility is located, in whole or in part, may require the owner or operator of the covered facility to prepare and provide facility impact reports as set forth in this section. A municipality that borders a municipality in which a covered facility is located, and that shares with the covered facility a water source, sewer system, or electric distribution infrastructure, may require that the facility impact reports prepared under this section be provided to it on the same terms.
(b) A facility impact report shall state, for the reporting period, drawn from readily available data and expressed in common, explainable units:
(1) the volume of water withdrawn by the facility and the volume of water consumed by the facility, stated separately, identified by source, together with the volume discharged to any sewer system;
(2) the electricity consumed by the facility, its peak demand, and the share of its consumption served by on-site generation, identified by generation type;
(3) the hours of operation of any backup or auxiliary generation, the fuel consumed by such generation, and the resulting emissions as calculated by the methods the owner or operator already employs in any report to a federal or state agency;
(4) whether information generated by the interactions of residents of the commonwealth with AI systems operated from the facility is retained at the facility, used to train or improve AI systems, or shared, stated in accessible and common language, to the extent known by the owner or operator, so that the value taken from a community is visible to the community itself; and
(5) any material change anticipated in the following reporting period in any matter reported under paragraphs (1) through (4).
(c) A municipality may require facility impact reports not more frequently than quarterly. The owner or operator shall provide each report to the municipality and shall publish each report so that it is accessible to the residents of the community. A report drawn from data the owner or operator already collects or already reports satisfies this section; nothing in this section requires material additional instrumentation, and nothing in this section requires the disclosure of proprietary algorithms or architectures, training data, model weights, or trade secrets, or any other confidential intellectual property.
(d) This section grants no authority to restrict, delay, condition, or prohibit the construction, operation, or use of any covered facility or any AI system, and none shall be inferred from it. It establishes a right to visibility, not a right to veto. The authority vested in the municipality by this section is in addition to, and does not diminish, any existing authority around land use, zoning, and public health.
(e) No state agency, executive order, regulation, or administrative action shall restrict, diminish, or preempt the authority of municipalities under this section. No provision of the general laws, whether enacted before or after the effective date of this section, shall be construed to restrict, diminish, or preempt the authority of municipalities under this section unless such provision specifically and explicitly states that it is intended to do so, and any ambiguity shall be resolved in favor of preserving such authority.
(f) Nothing in this section limits the right of residents to transparency around the outputs and inferences of any AI system that operates from a covered facility when used in consequential decisions around their lives, as provided in section 2.
(g) A municipality may seek injunctive relief to enforce the requirements of this section through appropriate proceedings in the superior court.
(h) A municipality in which a covered facility is proposed to be located, or is under construction, may require the person proposing or constructing the facility to provide a prospective facility impact report, stating the projected values of the matters described in subsection (b), drawn from the design specifications and interconnection filings for the facility and from readily available data, before the facility commences operation. A prospective report under this subsection is an instrument of visibility only; nothing in this subsection conditions any approval upon it, creates any approval process, or grants any authority to delay, restrict, or prohibit the construction or operation of the facility.
Section 5. State Procurement of AI Systems.
(a) This section shall apply to any enterprise AI deployment by the commonwealth or any state agency.
(b) No enterprise AI deployment shall proceed without a documented competitive evaluation of not fewer than 3 qualified vendors, conducted through a publicly accessible procurement process. The procurement process shall include:
(1) a public solicitation published on COMMBUYS or its successor system with not fewer than 45 days for vendor response;
(2) published evaluation criteria, including technical capability, security certification, data protection, cost, and interoperability;
(3) a written evaluation report comparing all responsive vendors against each published criterion;
(4) publication of the evaluation report, redacted only as required by law to protect genuine trade secrets, within 30 days of contract award; and
(5) submission by each responsive vendor of a truthful accounting of the AI system offered, including a statement of the conditions under which the vendor has evaluated the system and the conditions under which it has not. No vendor shall be deemed responsive without one. The truthful accounting of the vendor awarded the contract shall be published with the evaluation report under paragraph (4), and the capability of each offered system to disclose its use, its outputs, and its inferences to the people it is used upon shall be a published evaluation criterion under paragraph (2).
(c) Any individual serving on an advisory body, task force, evaluation committee, or other entity that recommends, evaluates, or influences the selection of an AI vendor for state procurement shall file a public disclosure identifying:
(1) any material financial relationship between the individual, or the individual's employer, and any prospective vendor;
(2) any campaign contributions made by the individual to any elected official with authority over the procurement within the preceding 48 months; and
(3) any board membership, advisory role, or consulting arrangement with any prospective vendor or the vendor's investors, partners, or parent company.
(d) Disclosures required under subsection (c) shall be filed with the office of the inspector general and made publicly available on the commonwealth's procurement transparency portal within 14 days of the individual's appointment or, if later, within 14 days of the commencement of any procurement process.
(e) An individual with a material financial relationship to a prospective vendor shall recuse themselves from any evaluation, recommendation, or vote pertaining to that vendor. Failure to disclose or recuse shall constitute a violation of chapter 268A and shall be referred to the state ethics commission.
(f) No enterprise AI deployment shall be implemented without a consultation period of not fewer than 60 days during which affected public employees and their designated representatives may review the proposed deployment, submit comments, and receive written responses to substantive concerns. Where a collective bargaining agreement covers affected employees, the deployment shall be subject to the applicable provisions of such agreement regarding the introduction of new technology.
(g) The procuring agency shall publish a workforce impact assessment identifying anticipated effects on job functions, required training, data access, and working conditions.
(h) No enterprise AI deployment contract shall exceed a term of 3 years without renewal. Renewal shall require a new competitive evaluation under the standards of this section.
(i) Every enterprise AI deployment shall include a quarterly performance review, the results of which shall be published and made available to the General Court, affected workers, and the public. Quarterly reviews shall address system performance, accuracy, observed drift as measured against the conditions of the evaluation conducted under subsection (b), fine-tuning or other material modifications since the prior review, and any issues identified by affected workers or residents.
(j) The General Court shall receive notice of any proposed enterprise AI deployment not fewer than 60 days prior to contract execution, including the proposed vendor, contract value, scope of deployment, and a summary of the competitive evaluation.
(k) The state auditor shall have authority to audit any enterprise AI deployment for compliance with the procurement standards established by this section. The inspector general shall have authority to investigate complaints regarding conflicts of interest, procurement irregularities, or violations of this section.
(l) Every enterprise AI deployment contract shall provide that the Authority may conduct, or direct the conduct of, inference audits of the deployed AI system. The results of each inference audit shall be published with the quarterly performance review under subsection (i), redacted only as required by law, and shall be provided to the Community Advisory Board.
(m) This section shall apply to enterprise AI deployment contracts entered into or renewed on or after the effective date of this act. Nothing in this section shall be construed to invalidate or require renegotiation of contracts entered into prior to said effective date.
(n) Exceptions. The requirements of subsections (b), (f), and (j) may be modified or waived only upon written findings by the secretary of administration and finance, published within 14 days and provided to the inspector general and the General Court, that: (1) fewer than 3 qualified vendors exist for the system sought, in which case the competitive evaluation shall include all qualified vendors; (2) an emergency affecting cybersecurity, public health, or public safety requires deployment before the periods prescribed can run, in which case any resulting contract shall not exceed 12 months and any renewal shall comply with this section in full; (3) the system is required by federal law or as a condition of federal funds; or (4) the deployment is a pilot of not more than 6 months' duration and $250,000 in value, not renewable except through full compliance with this section. Every waiver under this subsection shall be time-limited, shall be subject to audit under subsection (k), and shall not modify or waive the disclosure and publication requirements of this section.
Section 6. Community AI Understanding Program.
(a) The General Court finds that Massachusetts shall lead in the human-centered implementation of artificial intelligence, and that no statewide approach to AI can succeed without a grounded understanding of how AI is encountered, understood, and experienced by residents across the commonwealth's diverse communities.
(b) There is hereby established the Community AI Understanding Program, administered by the Authority in partnership with the Massachusetts Municipal Association as provided under section 11, through which municipalities may conduct annual assessments of how artificial intelligence affects their residents, institutions, and local economies. Participation in the program shall be voluntary.
(c) An annual Community AI Assessment may be conducted through any mechanism the municipality deems appropriate, including but not limited to a town meeting agenda item, a public forum, a resident survey, a report prepared by municipal staff or a designated appointee, or any combination thereof. Municipalities are encouraged to select assessment formats that reach residents whose civic participation patterns may not run through traditional municipal government channels. The assessment should address:
(1) how AI systems are currently encountered and experienced by residents of the municipality;
(2) the opportunities, concerns, knowledge gaps, and support needs of the community with respect to artificial intelligence;
(3) any consequential decisions affecting residents that are known to involve AI systems;
(4) the state of community understanding and readiness with respect to artificial intelligence; and
(5) the extent to which data generated by the interactions of residents with AI systems is known to be retained, used to train or improve such systems, or shared, and any facility impact reports provided under section 4 to the municipality during the assessment period.
(d) Municipalities are encouraged to submit completed assessments to the Community Advisory Board established in section 10 for inclusion in the statewide synthesis.
(e) Nothing in this section creates any regulatory authority, permitting requirement, review process, or mechanism by which a municipality may delay, condition, or restrict the deployment or use of any AI system. The Community AI Understanding Program is an instrument of awareness and understanding, not governance or enforcement.
(f) The Authority shall affirmatively recruit participation from geographically and demographically diverse municipalities, with specific attention to Gateway Cities, rural communities, and municipalities with significant populations for whom English is not the primary language at home. The Authority shall report annually on the geographic and demographic diversity of participating municipalities and shall take reasonable steps to address gaps in representation.
(g) The General Court shall appropriate such sums as may be necessary to support municipalities in conducting annual assessments, with priority given to municipalities with populations under 25,000 and municipalities with limited existing technical capacity. The Authority shall administer grants to participating municipalities from the Trust Fund established under section 13.
Section 7. AI Trust Certification, Innovation Sandbox, and Reciprocity.
(a) The General Court finds that transparency and innovation are mutually reinforcing. Jurisdictions that establish clear, predictable standards for AI transparency create conditions of trust that attract responsible investment, expand market access, and accelerate the development and deployment of trustworthy AI systems. The purpose of this section is to position the Commonwealth of Massachusetts as the premier jurisdiction for the development and deployment of transparent, trustworthy AI. The General Court intends that the transparency standards established by this chapter serve as a model for AI governance at the state and federal level and shall continue in full force unless expressly preempted by federal law.
(b) AI Trust Certification.
(1) There is hereby established the Massachusetts AI Trust Certification, a voluntary program administered by the Authority under which AI developers and deployers may earn a state-backed certification demonstrating compliance with transparency and accountability standards.
(2) The AI Trust Certification program shall:
(A) establish tiered certification levels reflecting the degree and scope of transparency and accountability demonstrated by the applicant, including at minimum a foundational tier based on self-attestation to recognized standards, an intermediate tier based on third-party verification, and an advanced tier based on internationally recognized management-system certification;
(B) where appropriate, reference and incorporate existing recognized frameworks, including but not limited to the National Institute of Standards and Technology AI Risk Management Framework, ISO/IEC 42001, and comparable international standards, to promote interoperability;
(C) publish certification criteria, application procedures, and evaluation standards in advance and make them freely available to prospective applicants;
(D) conduct evaluations through a transparent review process, with results published within 60 days of application;
(E) issue certifications valid for a period of 2 years, renewable through demonstrated continued compliance;
(F) specify the scope of each certification, including the specific AI systems and deployment contexts covered, and require certified entities to use the certification mark only in connection with that specified scope; and
(G) maintain a public registry of all certified entities, including the tier of certification, the scope of certification, the date of issuance and expiration, and the status of any open formal inquiries or revocation proceedings under section 10.
(3) A certified entity shall notify the Authority within 30 days of any material modification to a certified AI system. The Authority shall by regulation establish the criteria for material modification, procedures for review, and timelines for re-evaluation, which shall be proportionate to the risk and scope of the modification.
(c) Entities holding a current Massachusetts AI Trust Certification shall be eligible for:
(1) preferred consideration in state and municipal AI procurement processes, provided such preference does not override the competitive evaluation requirements of section 5 of this chapter;
(2) eligibility for innovation grants funded through appropriations designated for the purposes of this section or through the Trust Fund;
(3) expedited transparency compliance within participating municipalities, under which a current certification may be presented as presumptive evidence of compliance with municipal transparency requirements, to the extent such requirements fall within the scope of the certification; and
(4) use of the Massachusetts AI Trust Certification mark in commercial materials, subject to terms established by the Authority, provided that the mark may be used only in connection with the specific AI systems and deployment contexts within the scope of the certification, and any use of the mark shall disclose the tier and scope.
(d) Innovation Sandbox.
(1) There is hereby established an Innovation Sandbox through which entities may deploy experimental AI systems in participating municipalities under enhanced transparency conditions. The Innovation Sandbox shall:
(A) permit accelerated deployment timelines for AI systems that are not yet eligible for full AI Trust Certification, provided the deploying entity agrees to enhanced transparency and community feedback mechanisms exceeding the standard requirements of this chapter;
(B) limit experimental deployments to a period of not more than 12 months, renewable once upon demonstration of satisfactory performance and community benefit, and subject to review by the Community Advisory Board;
(C) require the informed consent of the participating municipality's governing body;
(D) require the deploying entity to publish quarterly public reports detailing system performance, identified issues, community feedback received, and corrective actions taken;
(E) provide a structured pathway from sandbox participation to full AI Trust Certification; and
(F) require public disclosure of the reason for any exit from the sandbox without successful transition to full certification.
(2) The Authority shall by regulation establish application procedures, participation criteria, data-usage limits, required consumer disclosures, reporting requirements, and fees for the Innovation Sandbox. Such regulations shall include substantive review of sandbox applications by the Community Advisory Board and a requirement that the Authority respond in writing to Board recommendations regarding sandbox admissions and exits.
(e) Reciprocity.
(1) The Authority shall actively seek reciprocity agreements with other states, regional compacts, federal agencies, and international jurisdictions that maintain AI transparency frameworks compatible with the standards established under this chapter. Reciprocity agreements may provide for:
(A) mutual recognition of AI trust certifications or equivalent credentials;
(B) shared technical standards for transparency and interoperability;
(C) coordinated enforcement of AI transparency requirements; and
(D) coordinated approaches to emerging AI transparency challenges.
(2) Reciprocity agreements shall be approved by the Authority's governing board following equivalency analysis and consultation with the Community Advisory Board. No mutual recognition shall be extended to a partner jurisdiction unless the Authority finds that the partner jurisdiction's standards meet the floor established by this chapter.
(3) An entity that has been denied Massachusetts certification on substantive grounds shall be ineligible for reciprocity-based operation in the commonwealth for a period to be established by the Authority by regulation.
(f) Annual Innovation and Trust Report. The Authority shall publish an annual Innovation and Trust Report assessing the effectiveness of the AI Trust Certification, Innovation Sandbox, and reciprocity programs, including: the number of certifications issued, renewed, suspended, and revoked; the scope and tier distribution of certifications; the economic impact of certified entities within the commonwealth; the outcomes of sandbox deployments; the status of reciprocity agreements; the status of AI transparency legislation in other states and at the federal level; and recommendations for maintaining the commonwealth's leadership position. The report shall be submitted to the General Court, the Community Advisory Board, and the Massachusetts Municipal Association, and shall be made publicly available.
Section 8. Legislative Oversight.
(a) The Joint Committee on Advanced Information Technology, the Internet and Cybersecurity, or such successor committee as the General Court may designate, shall exercise oversight of this chapter, including:
(1) receiving and reviewing the annual reports of the Authority, the Community Advisory Board, and the Massachusetts Municipal Association regarding its activities under section 11;
(2) reviewing compliance with the procurement standards established in section 5;
(3) monitoring any state or federal legislative proposals that would affect municipal AI transparency authority;
(4) recommending amendments as necessary to address emerging technologies, governance challenges, or threats to transparency rights; and
(5) receiving referrals from the Community Advisory Board under section 10(k) regarding disputes between the Board and the Authority or concerns regarding the systematic performance of the Authority.
(b) The Joint Committee shall hold at least 1 public hearing annually to receive testimony from the Authority, the Community Advisory Board, the Massachusetts Municipal Association, municipalities, affected workers, and members of the public regarding the implementation and effectiveness of this chapter.
(c) Upon receipt of a referral from the Community Advisory Board under section 10(k), the Joint Committee shall hold a public hearing within 90 days at which the Executive Director of the Authority and the Chair of the Community Advisory Board shall appear and respond.
Section 9. Massachusetts AI Trust Authority.
(a) Establishment. There shall be established within the executive office for administration and finance, but not under its control, a state agency known as the Massachusetts AI Trust Authority. The Authority shall be an independent public entity not subject to the supervision and control of any other executive office, department, commission, board, bureau, agency or political subdivision of the commonwealth, and shall have all of the powers and duties set forth in this chapter.
(b) Governing board.
(1) The Authority shall be governed by a board of 13 members, of whom 11 shall be voting members, consisting of:
(A) the secretary of economic development or designee, ex officio, non-voting;
(B) 3 members appointed by the governor, of whom one shall serve as chairperson;
(C) 2 members appointed by the attorney general;
(D) 2 members appointed by the state auditor;
(E) 1 member appointed by the president of the senate;
(F) 1 member appointed by the speaker of the house of representatives;
(G) 2 members selected by the Community Advisory Board from among its seated members, each of whom shall serve on the governing board for the duration of their service on the Community Advisory Board and shall not vote on any revocation recommendation in which they participated as a member of the Community Advisory Board; and
(H) 1 member designated by the Massachusetts Municipal Association as provided in section 11, non-voting.
(2) Appointed members shall serve staggered 5-year terms. Initial appointments shall be designated by the appointing authorities to create a staggered schedule such that no more than 3 appointed seats are subject to replacement in any single year.
(3) Members shall include, collectively, persons with demonstrated expertise in machine learning or artificial intelligence engineering, information security, civil rights and consumer protection, organized labor, municipal governance, academic AI research, and industry deployment of AI systems. At least one appointed member shall be a resident of a Gateway City.
(4) No appointed member shall hold any other full- or part-time employment in the executive branch of the commonwealth during service. All appointed members shall be deemed special state employees for purposes of chapter 268A. During service and for 1 year following service, no appointed member shall be employed by, consult for, or hold equity in any entity certified by the Authority, applying for certification, or participating in the Innovation Sandbox.
(5) No individual with a material financial relationship to any AI vendor holding or seeking a Massachusetts AI Trust Certification or Innovation Sandbox admission shall be eligible for appointment to the governing board.
(6) A majority of seated voting members shall constitute a quorum. Adoption or amendment of certification criteria, sandbox standards, or reciprocity agreements, and action on revocation recommendations from the Community Advisory Board, shall require the affirmative vote of not fewer than 8 voting members.
(c) Powers and duties. The Authority shall have the power to:
(1) adopt regulations governing voluntary AI Trust Certification, the Innovation Sandbox, reciprocity agreements, material-modification review, delegated-transaction thresholds, and other matters within its jurisdiction;
(2) establish by rule application procedures, participation criteria, data-usage limits, required consumer disclosures, reporting requirements, and fees for all programs administered by the Authority;
(3) conduct investigations, inference audits, and public hearings, including the inference audits provided for in section 5(l);
(4) enter into reciprocity and mutual-recognition agreements with other states, federal agencies, and international bodies;
(5) designate accredited third-party assessors to conduct evaluations on behalf of the Authority, subject to standards established by regulation;
(6) accept appropriations, federal funds, grants, and private gifts;
(7) make contracts and execute instruments necessary to its functions, including contracts with the Massachusetts Municipal Association, the Collins Center for Public Management at the University of Massachusetts Boston, and other qualified entities for specified community engagement, municipal coordination, and technical assistance functions;
(8) employ staff outside the civil service system and set compensation appropriate to specialized technical talent; and
(9) publish information, reports, and advisory opinions, without the approval of any other officer or employee of any executive agency with respect to the substance of any report or publication.
(d) Executive Director. The governing board shall appoint an Executive Director by majority vote. The Executive Director shall serve at the pleasure of the board and shall have responsibility for the day-to-day operations of the Authority, including the deployment of staff resources consistent with priorities established by the governing board and, with respect to Community Advisory Board operations, by the Community Advisory Board.
(e) Post-employment restrictions.
(1) No staff member of the Authority involved in certification decisions, sandbox admissions, or investigations may, for a period of 1 year following separation from the Authority, be employed by, consult for, or hold equity in any entity with respect to which the staff member made certification decisions, sandbox decisions, or investigation-related decisions within the preceding 24 months.
(2) The Authority shall by regulation establish procedures for monitoring and enforcing the post-employment restrictions of this subsection.
(f) Trade-secret confidentiality. Documentary materials or data consisting of trade secrets or confidential technical, commercial, or financial information of entities participating in the Authority's voluntary certification program, Innovation Sandbox, or reciprocity program shall not be public records within the meaning of section 10 of chapter 66, or clause Twenty-sixth of section 7 of chapter 4. Nothing in this subsection limits the obligation of the Authority to maintain a public registry of certified entities and certification status as required by section 7(b)(2)(G).
(g) Conflict of interest. Members of the governing board and staff of the Authority are subject to chapter 268A. The Authority shall by regulation establish additional ethics standards appropriate to the specialized nature of its work.
Section 10. Community Advisory Board.
(a) Establishment and purpose. There is hereby established within the Authority a Community Advisory Board to synthesize community-level understanding of artificial intelligence into a statewide picture that informs policy, innovation, and democratic participation, and to exercise the oversight, petition, investigatory, and revocation-recommendation authorities set forth in this section. The Community Advisory Board shall be a statutorily independent body within the Authority, funded through the Trust Fund, and not subject to direction by the Executive Director of the Authority with respect to the exercise of its authorities under this section.
(b) Composition. The Community Advisory Board shall consist of 11 voting members, together with 1 non-voting member designated by the Massachusetts Municipal Association as provided in section 11. The voting members shall be selected as follows:
(1) 7 members nominated by the governing bodies of municipalities that have participated in the Community AI Understanding Program and selected through the public application and review process described in subsection (e);
(2) 1 member representing organized labor, nominated by the Massachusetts AFL-CIO and confirmed through the public review process described in subsection (e);
(3) 1 member representing civil liberties and digital rights organizations, nominated on a rotating basis among organizations designated by the Authority in consultation with the Massachusetts Municipal Association for this purpose, with no single organization holding the nomination right for more than 1 consecutive term, and confirmed through the public review process described in subsection (e); and
(4) 2 at-large members selected through open public application and review by the seated members of the Community Advisory Board, with preference for applicants demonstrating expertise in data governance, community organizing, municipal administration, or the effects of automated systems on underserved populations.
(c) Diversity and expertise requirements.
(1) The municipal seats under subsection (b)(1) shall not be more than 60 percent drawn from any single region of the commonwealth or from any single category of municipality defined by size or demographic composition. The Authority shall maintain standards to ensure geographic and demographic diversity among municipal seats.
(2) The Community Advisory Board shall include, collectively, not fewer than 2 members with demonstrated technical expertise in artificial intelligence, machine learning, or algorithmic systems, of whom not more than 1 shall be employed by a for-profit entity in the AI industry and at least 1 shall hold an academic or independent position. This requirement may be satisfied through the at-large seats, the municipal seats, or any combination thereof.
(d) Terms of service and leadership.
(1) Each member shall serve a term of 3 years.
(2) No member shall serve more than 2 consecutive terms. A member who has served 2 consecutive terms shall be ineligible for reappointment until the expiration of 1 full term following the conclusion of their service.
(3) Terms shall be staggered so that no more than 4 seats are subject to selection in any single year. For initial appointments, the selecting bodies shall designate 3 members for initial terms of 3 years, 4 members for initial terms of 2 years, and 4 members for initial terms of 1 year.
(4) A vacancy occurring before the expiration of a term shall be filled through the same public application and selection process for the remainder of the unexpired term. Service to fill a vacancy of fewer than 18 months shall not count toward the 2-term limit.
(5) The Board shall annually elect from among its seated members a Chair and Vice-Chair, each to serve 1-year renewable terms. The Chair shall not be in their final year of service. The Vice-Chair shall be at a different point in the term cycle than the Chair to ensure leadership continuity across turnover.
(6) A member may be removed for cause, including violation of confidentiality obligations, failure to disclose conflicts of interest, abuse of investigatory authority, or conduct incompatible with Board service, upon a two-thirds vote of the seated Board with cause documented in the public record. Removal is subject to appeal to the Superior Court.
(e) Public application and selection process.
(1) All nominations and applications for the Community Advisory Board shall be submitted through a public application process administered by the Authority in partnership with the Massachusetts Municipal Association.
(2) Applications shall be publicly available for review for a period of not fewer than 30 days prior to selection.
(3) Selection criteria shall be published in advance, including but not limited to: demonstrated commitment to community engagement, relevant expertise or lived experience, geographic and demographic diversity, and absence of material financial relationships with AI vendors holding or seeking state or municipal AI contracts or certifications.
(4) The Authority shall publish the names and qualifications of all applicants, a written summary of the review process, and the basis for each selection decision within 14 days of final selection; provided that home addresses, personal contact information, and other personal identifying information of applicants shall be redacted from publication.
(f) Conflict of interest.
(1) No individual with a material financial relationship to any AI vendor holding or seeking a Massachusetts AI Trust Certification or a state or municipal AI contract shall be eligible to serve on the Community Advisory Board; provided, however, that for purposes of eligibility for the seat described in subsection (c)(2), employment by a for-profit entity in the AI industry shall not, standing alone, constitute a material financial relationship unless the employer holds or is seeking a Massachusetts AI Trust Certification, Innovation Sandbox admission, or a state or municipal AI contract.
(2) An individual with pending litigation against, or pending employment negotiation with, any such entity shall disclose that fact upon application and upon its arising, and shall not participate in any Board action concerning that entity during its pendency.
(3) No individual who has made public statements advocating the shutdown, prohibition, or regulatory sanction of a specific identified entity within the preceding 24 months shall participate in any Board action concerning that entity, and shall disclose such statements upon seating.
(4) A member nominated by a municipality that has received material funding from a certified entity, a sandbox participant, or an applicant for certification or sandbox admission within the preceding 36 months shall disclose that relationship publicly, and shall not participate in any Board action concerning the funding entity during the pendency of the relationship and for 24 months thereafter.
(5) Members shall annually file public disclosures identifying any relationships that could reasonably give rise to a conflict of interest under this subsection.
(g) Core duties and authorities. The Community Advisory Board shall:
(1) synthesize the findings of municipal Community AI Assessments into a statewide understanding of how artificial intelligence is experienced across the diverse communities of the commonwealth, identifying both the common values and the distinct local conditions that must inform state policy;
(2) publish an annual report on the state of human understanding and readiness with respect to artificial intelligence across the commonwealth, including an independent assessment of the Authority's performance and whether certification and sandbox programs are reaching the communities the Board represents;
(3) develop and publish recommended standards and plain-language guides to assist residents in understanding and exercising the rights established by section 2, including guides translated into the principal languages spoken across the commonwealth;
(4) recommend to the Authority, on a rolling basis, proposed additions to the categories of consequential decision under section 1 that the Authority may adopt by regulation consistent with the standard established by that section, and recommend to the Joint Committee on Advanced Information Technology, the Internet and Cybersecurity proposed amendments to this chapter to address AI applications that materially affect residents but fall outside the scope of existing categories or the Authority's regulatory authority; and
(5) advise the General Court, the governor, the Authority, and the Massachusetts Municipal Association on matters affecting communities with respect to artificial intelligence.
(h) Investigatory authority.
(1) The Community Advisory Board may initiate formal inquiries into whether certified entities, sandbox participants, or applicants are complying with their certification or sandbox terms, with the transparency obligations of section 2, or with the disclosure obligations of section 5, upon substantiated complaint from an affected resident, a municipality, a public employee, or a designated representative.
(2) A formal inquiry shall be opened only by the affirmative vote of not fewer than two-thirds of the seated voting members, and in no event fewer than 6, with the vote and rationale documented in the public record. An inquiry shall not be initiated or pursued at the direction of any individual member without such Board action.
(3) In the course of a formal inquiry, the Board may require the entity under inquiry to provide information relevant to the inquiry. The Board may seek enforcement of such requirements through subpoena issued in accordance with procedural rules governing the Board's own proceedings, which the Board shall adopt and publish and which shall provide, at minimum, standards of relevance and proportionality of scope, the preservation of applicable privileges, protection of trade secrets and confidential information consistent with section 9(f), reasonable methods of service, an opportunity to object, and modification or quashal by the Superior Court. A subpoena under this paragraph shall be enforceable through proceedings in the Superior Court.
(4) A formal inquiry shall conclude within 18 months of initiation unless extended by the affirmative vote of not fewer than two-thirds of the seated voting members, and in no event fewer than 6, with published reasons. The Board shall publish findings at the conclusion of each inquiry, including findings of no violation where applicable.
(5) Upon finding substantial evidence of non-compliance, the Board may refer the matter to the attorney general, the inspector general, the state ethics commission, or the governing board of the Authority for appropriate action.
(6) Certified entities, sandbox participants, and applicants shall not be required to respond to informal inquiries by individual Board members. Only formally opened inquiries pursuant to paragraph (2) carry any obligation to respond.
(7) The Board may prohibit, by rule, any funding, coordination, or organization by certified entities, sandbox participants, or applicants of community feedback submitted to the Board. Undisclosed activity of this nature shall constitute grounds for revocation recommendation.
(i) Petition authority.
(1) The Community Advisory Board may formally petition the governing board of the Authority to adopt, amend, or rescind any regulation, certification criterion, sandbox standard, or reciprocity agreement.
(2) The governing board of the Authority shall respond in writing to any such petition within 60 days, addressing the substantive concerns raised and stating the basis for its action or inaction.
(3) Prior to adopting or amending certification criteria, sandbox admission standards, or reciprocity agreements, the governing board of the Authority shall submit the proposed action to the Community Advisory Board for review and comment over a period of not fewer than 60 days and shall publish its written response to substantive comments alongside any final adoption.
(4) If the governing board of the Authority fails to respond to a petition within the required time, or if its response does not address the substantive concerns raised in the petition, the Community Advisory Board may seek declaratory judgment in the Superior Court regarding the Authority's compliance with this subsection.
(j) Revocation recommendation authority.
(1) The Community Advisory Board may formally recommend revocation, suspension, or modification of any Massachusetts AI Trust Certification or sandbox participation. A formal recommendation shall be made only by the affirmative vote of not fewer than two-thirds of the seated voting members, and in no event fewer than 6, with the vote and rationale documented in the public record.
(2) The governing board of the Authority shall vote publicly on any revocation recommendation within 90 days of receipt. If the governing board declines to act on the recommendation, it shall publish its written reasoning.
(3) Voluntary withdrawal from certification or sandbox participation during the pendency of a formal inquiry or revocation proceeding shall not moot the proceeding. The Authority shall proceed to a finding on whether revocation would have been appropriate and shall note any such voluntary withdrawal in the public registry.
(k) Escalation to legislative oversight. If the Community Advisory Board, by the affirmative vote of not fewer than two-thirds of the seated voting members, and in no event fewer than 6, concludes that the Authority has systematically failed to fulfill its statutory mission or has unreasonably failed to respond to the Board's petitions or recommendations, the Board may refer the matter to the Joint Committee on Advanced Information Technology, the Internet and Cybersecurity under section 8(a)(5).
(l) Limitation of authority. The Community Advisory Board shall have no regulatory authority, no permitting authority, and no authority to directly delay, condition, or restrict the deployment or use of any AI system. The authorities granted under this section are authorities of inquiry, petition, recommendation, and referral. Final action with respect to certification, sandbox admission, or reciprocity rests with the governing board of the Authority. The Board's jurisdiction extends to certified entities, sandbox participants, and applicants; it does not extend to the internal operations or personnel decisions of the Authority.
(m) Rights of entities under inquiry.
(1) An entity that is the subject of a formal inquiry has the right to notice of the inquiry, a statement of the matters under inquiry, and a reasonable opportunity to respond.
(2) At the conclusion of an inquiry in which no violation is found, the Board's findings shall clearly state this result, and the entity shall have the right to a public statement in the Board's findings.
(3) An entity subjected to a formal inquiry that a court finds to have been conducted without adequate factual basis may seek declaratory relief and recovery of reasonable legal costs. This remedy shall not be available for inquiries that identified partial violations, concerning practices short of violations, or conduct warranting further review.
(n) Confidentiality and member conduct.
(1) Board members serve under a confidentiality obligation regarding non-public information obtained through Board proceedings. Leaks or unauthorized disclosures of non-public information shall constitute grounds for removal under subsection (d)(6).
(2) Public statements by individual members about pending Board matters are prohibited. Only the Board's formal published findings and actions speak for the Board. This paragraph does not restrict individual members' First Amendment rights to speak about general AI policy matters not directly before the Board.
(3) Board members shall not communicate with certified entities, sandbox participants, or applicants about pending matters outside formal Board proceedings. Any such communications shall be disclosed to the Board and, upon conclusion of the matter, to the public.
(4) Board members who declare candidacy for partisan elected office shall resign from the Board immediately. The Board's prior work involving matters in which the former member substantively participated shall be subject to retroactive review for potential conflicts.
(o) Meetings. Community Advisory Board meetings shall be open to the public and subject to the open meeting law, sections 18 to 25, inclusive, of chapter 30A of the General Laws. Meeting agendas, minutes, and materials shall be published on a public website not fewer than 48 hours before each meeting and within 7 days following each meeting, respectively. The Board may hold non-public executive sessions solely for the purpose of discussing specific matters of ongoing inquiry that would be prejudiced by public discussion, consistent with the executive session provisions of section 21 of said chapter 30A.
(p) Resources.
(1) The Community Advisory Board shall have dedicated professional staff support within the Authority. The Authority shall provide the Board with not less than the equivalent of 5 full-time professional staff positions at commencement of Board operations, increasing as caseload warrants, including legal, investigative, technical, and policy staff. Board staff shall report to the Board with respect to the exercise of the Board's authorities under this section.
(2) Board members shall receive an annual stipend of not less than $15,000 in recognition of the time commitment required, together with reimbursement for expenses reasonably incurred in Board service.
(3) The Board shall have access to independent technical expertise through contract or consulting arrangements, with such access not contingent on the approval of the Executive Director of the Authority.
(4) The Board's annual operating budget shall be not less than 15 percent of the Authority's annual operating budget, drawn from the Trust Fund. This floor shall not be construed to limit the Board's ability to request or receive additional resources as warranted.
(5) Newly seated Board members shall receive substantive technical orientation, arranged by the Authority in consultation with academic institutions or independent experts, during the first 120 days of their term.
Section 11. The Relationship with the Massachusetts Municipal Association.
(a) Named consultation partner. The MMA is recognized as a named consultation partner of the Authority when dealing with matters impacting municipalities. Before adopting or amending any rule that impacts municipal involvement in the Community AI Understanding Program, the Innovation Sandbox, reciprocity agreements or municipal access to AI Trust Certification processes, the Authority shall seek input from or consult with the MMA.
(b) Partnership functions. The Authority has the right to engage the MMA through contract or intergovernmental agreement to perform any of the following functions on behalf of the Authority or to support any of the programs administered by the Authority:
(1) administration of municipally related peer learning networks, training for officials and the development and publication of model municipal policies;
(2) technical assistance rendered to small towns defined as having populations under 25,000 in support of participation in the Community AI Understanding Program;
(3) incident reporting that shall aggregate said municipal AI incidents up to the Authority and the Community Advisory Board; and
(4) support function for recruiting geographically and/or demographically diverse municipalities for the Authority's obligation under section 6(f).
(c) Guaranteed representation. One non-voting seat will be held by the MMA on both the Community Advisory Board and the Authority's governing board, as provided in sections 9(b)(1)(H) and 10(b), which shall be for the purpose of ensuring efficient and smooth cross-communication between the MMA and both of said bodies.
(d) Deemed compliance. Model municipal AI policies described in subsection (b)(1) of this section, adopted by a municipality and written and published by the MMA without any materially relevant changes, shall presumptively satisfy any baseline municipal requirements set forth by the Authority, subject to review by the Community Advisory Board.
(e) Funding. The MMA shall be reimbursed by the Authority for performing the functions described under subsection (b) through a budget line from the Authority and drawn from the Trust Fund. The Authority and the MMA will together draft an agreement defining the level of funding that shall be sufficient to support all the functions assigned to the MMA.
(f) Alternative partners. Functions under subsection (b) that the MMA declines to perform, or where the Authority decides another partner is better equipped or suited to perform, the Authority may choose to contract with the Collins Center for Public Management at the University of Massachusetts Boston or any other qualified entity subjected to a competitive process.
(g) Non-displacement. Nothing in this section shall be construed as a displacement, restriction or otherwise a diminishment of the independent advocacy role of the MMA on behalf of its municipal membership and with respect to this chapter or any other matter regardless of nature.
Section 12. Intermunicipal Cooperation.
(a) In order to build shared infrastructure concerning transparency, for conducting joint Community AI Assessments and for the pooling of resources relating to data governance and AI oversight capacity, the municipalities of the commonwealth may voluntarily enter into cooperative agreements.
(b) Data generated by or through a cooperation framework entered into under this section shall be owned exclusively by the municipality within whose jurisdiction it was generated, and that municipality shall retain primary authority over said data, which is affirmed by the General Court.
(c) From the Trust Fund the Authority shall administer grants to support voluntary municipal cooperation with priority to be given to small municipalities defined as municipalities with populations under 25,000 and/or with limited technical capacity to be determined by the Authority and the MMA.
Section 13. Massachusetts AI Trust Fund.
(a) Trust Fund. There shall be established and placed within the Authority a fund to be known as the Massachusetts AI Trust Fund, to be held by the Authority separate and apart from its other funds. The Fund shall be credited with:
(1) any appropriations, bond proceeds, federal grants, or other monies authorized to be credited thereto by the General Court;
(2) fees established by the Authority by regulation for AI Trust Certification applications, renewals, Innovation Sandbox participation, reciprocity processing, and related services;
(3) assessments established by the Authority under subsection (c);
(4) grants from foundations, federal agencies, and other non-state sources received for purposes consistent with this chapter; and
(5) any other monies which may be available and are directed to the Fund.
(b) Expenditure. The Authority shall expend monies in the Fund for the purposes of this chapter without further appropriation, including for operation of the Authority, the Community Advisory Board, programs administered by the Authority, grants to municipalities under sections 6 and 12, reimbursement to the Massachusetts Municipal Association under section 11, and other purposes consistent with this chapter.
(c) Industry assessment.
(1) The Authority may annually assess covered AI deployers and developers doing business in the commonwealth, in such amounts and on such formula as the Authority shall determine by regulation, to fund the operations of the Authority and its programs.
(2) The Authority shall establish by regulation the categories of deployers and developers subject to assessment, reasonable thresholds for assessment based on Massachusetts-related AI revenue or scale of AI deployment, and a formula designed to distribute the assessment burden equitably across categories of regulated entities; provided that no entity with annual Massachusetts-related AI revenue below $10,000,000 shall be subject to assessment, and provided further that the categories initially subject to assessment shall be limited to certified entities, sandbox participants, applicants for certification or sandbox admission, and vendors under enterprise AI deployment contracts with the commonwealth.
(3) No assessment shall exceed the amount reasonably necessary to fund the Authority's operations and programs, taking into account other available funding sources; the aggregate of all assessments in any fiscal year shall not exceed the Authority's approved operating budget for that year; and each assessment shall bear a reasonable relationship to the cost of the regulatory and programmatic services provided by the Authority to the entities assessed, it being the intent of the General Court that the assessment operate as a regulatory fee and not as a tax.
(4) The Authority shall publish the methodology by which each assessment is calculated and shall provide by regulation an administrative process through which an assessed entity may seek review of the calculation or applicability of an assessment, with further review available under chapter 30A.
(d) Fee schedule. The Authority shall establish by regulation a fee schedule for AI Trust Certification applications, renewals, Innovation Sandbox participation, and reciprocity processing. The fee schedule shall be tiered by the size of the applying entity and the tier of certification sought, and shall include reduced fees for startups, small businesses, and organizations with limited capacity. Fees shall be reviewed not less than every 3 years and adjusted as necessary to maintain proportionality to the cost of providing the relevant services.
(e) Audit. The Authority shall be subject to audit by the state auditor. The Authority shall publish audited financial statements annually.
Section 14. Limitations.
(a) Except for the disclosure, recordkeeping, reporting, auditing, procurement, certification, and facility impact reporting obligations expressly established by this chapter, nothing in this chapter shall be construed to:
(1) prohibit, restrict, condition, license, or impose operational requirements upon the development, training, deployment, or use of any artificial intelligence system; provided, however, that the facility impact reporting authorized by section 4 concerns the physical operation of covered facilities and does not constitute an operational requirement upon any AI system within the meaning of this paragraph;
(2) create any regulatory authority, permitting requirement, or approval process for AI systems at the state or municipal level;
(3) require the disclosure of proprietary algorithms, trade secrets, model weights, training data, or other confidential intellectual property; the disclosure obligations of section 2 attach solely to the application of the AI system, at inference, to the affected individual, and shall not be construed to require any examination of or inquiry into training data, training methodology, or model development processes;
(4) require the documentation, inventory, or auditing of training data used to develop foundation models or large language models, the General Court having found that, as of the effective date of this Act, such requirements are technically infeasible for these systems using currently available methods and do not address the inference-time risks that are the subject of this chapter; provided that nothing in this subsection limits the authority of existing law to require training data documentation for narrow AI systems where such documentation is feasible; and provided further that this subsection shall not be construed to limit the authority of the General Court to revisit training data documentation requirements as methods and technical standards evolve;
(5) impose liability based solely on the composition or provenance of training data for foundation models or large language models, the General Court having determined that the relevant transparency obligation for these systems attaches at the point of impact on individuals, not at the point of model development; or
(6) limit the right of any person or entity to develop, train, deploy, or use computational resources for lawful purposes; the visibility provided in section 4 is not a limitation of that right.
(b) This chapter establishes transparency at the point of impact. It requires that when AI systems make or substantially inform consequential decisions about Massachusetts residents, those residents can see it. It extends the same principle to delegated transactions in which AI systems act autonomously on behalf of residents, and to the community impacts of the physical infrastructure from which these systems operate. It does not dictate how AI systems are built, what data they are trained on, or how they operate internally. The Commonwealth of Massachusetts welcomes the development and deployment of artificial intelligence and seeks to create conditions of trust and visibility in which innovation thrives.
(c) Nothing in the general laws shall be construed to diminish municipal AI transparency authority as affirmed by sections 3 and 4 of this chapter except by express legislative statement specifically and explicitly identifying its intent to restrict, diminish, or preempt such authority.
(d) Nothing in this chapter shall be construed to limit any right, remedy, or protection available to any resident, worker, or municipality under any other provision of the general laws, federal law, or municipal ordinance.
SECTION 4. Severability.
If any provision of this act or the application thereof to any person or circumstance is held invalid, such invalidity shall not affect other provisions or applications of this act which can be given effect without the invalid provision or application, and to this end the provisions of this act are declared to be severable.
SECTION 5. Effective Date and Transition.
(a) This act shall take effect on January 1, 2028, except as otherwise provided in this section.
(b) The Authority established under section 9 of the chapter inserted by section 4 of this act may take such preparatory actions as necessary prior to the effective date, including appointing an initial governing board, hiring initial staff, developing regulations, establishing AI Trust Certification criteria, soliciting municipal participation in the Community AI Understanding Program, convening the initial Community Advisory Board selection process, and entering into initial consultation with the Massachusetts Municipal Association under section 11 of said chapter.
(c) The appointing authorities for the governing board of the Authority shall make initial appointments not later than 180 days following the enactment of this act. The initial Community Advisory Board selection process shall conclude not later than 360 days following the enactment of this act.
(d) Prior to the seating of the members selected by the Community Advisory Board under section 9(b)(1)(G) of said chapter, the governing board of the Authority may organize and exercise all powers of the Authority with the seats then filled, and a majority of the seated voting members shall constitute a quorum for that period.
(e) For the initial selection of the Community Advisory Board only, the municipal seats under section 10(b)(1) of said chapter may be filled upon the nomination of the governing body of any municipality that has, by vote of its governing body, committed to participate in the Community AI Understanding Program.
(f) Regulations adopted by the Authority under said chapter shall be subject to the provisions of chapter 30A of the general laws governing the adoption of regulations by state agencies.
The information contained in this website is for general information purposes only. The General Court provides this information as a public service and while we endeavor to keep the data accurate and current to the best of our ability, we make no representations or warranties of any kind, express or implied, about the completeness, accuracy, reliability, suitability or availability with respect to the website or the information contained on the website for any purpose. Any reliance you place on such information is therefore strictly at your own risk.