- jurisdiction
- US-MA
- level
- state
- domain
- safety
- status
- proposed
- scope
- Frontier-AI safety requirements embedded in the Massachusetts FY economic development package (S.3178, an amended version of H.5576), requiring large frontier AI developers to publish safety frameworks addressing catastrophic and critical-safety-incident risk.
- obligations
- Large frontier AI developers must set and publish safety frameworks that mitigate their models' risk of catastrophes and critical safety incidents; the bill establishes a commission to study further AI regulation, including possibly requiring independent third-party audits and compliance verification; the Attorney General is empowered to bring civil actions against large frontier developers that violate the safety requirement.
- penalties
- Civil enforcement action by the Massachusetts Attorney General against noncompliant large frontier AI developers; specific monetary penalty amounts are not stated in the available legislative materials.
- appliesTo
- Large frontier AI model developers
- notes
- The House's own version of the economic-development bill stripped the frontier-AI safety language twice before the Senate acted (Economic Development/Ways and Means redraft removed it; a floor amendment, #166, to restore it was rejected). The Senate then passed S.3178, its amendment to H.5576, with frontier-AI provisions included, with floor debate concluding July 23, 2026, and sent it back to the House for further consideration; the House non-concurred and both chambers appointed conference committees on July 30, 2026. UPDATE (re-verified 2026-09-07 directly against malegislature.gov's own action histories for both H.5576 and S.3178): no further action has been recorded on either bill's official history since July 30, 2026. H.5576's own action history's most recent entries are all dated 7/30/2026: 'House NON-concurred in the Senate amendment' (this exact string reproduced verbatim across repeated independent fetches of the page), followed the same day by both the House and Senate appointing conference committees. No entry of any kind appears on the bill's page after that date as of this check. Secondary reporting (Boston Globe's 'An AI fight hits the Massachusetts Senate' and Insurance Journal, both through August 20, 2026) corroborates that the six-member conference committee was still negotiating the AI provisions as of that date, with Governor Healey not having committed to including the Senate's AI safeguards in her own economic-development proposal, and the House still resistant to language it had twice rejected. No source located confirms a conference report, a floor vote on a conference product, or gubernatorial signature/veto as of 2026-09-07. This remains an unresolved, unsettled bill: it has not been enacted and there is no confirmed outcome (survived, weakened, or dropped) for the frontier-AI safety language. Checked this round to settle the open item: malegislature.gov/Bills/194/S3178 (full action history to date, primary), malegislature.gov/Bills/194/H5576 (full action history to date, primary), plus secondary coverage (Boston Globe, Insurance Journal, MASSterList) used only to corroborate the primary-source silence, not to substitute for it. Status remains 'proposed' pending an actual conference report and enactment; the next run should re-check both bill pages directly rather than re-deriving from search, since informal/formal session rules mean action could resume without much advance notice. Distinct from the already-tracked frontier-AI-safety laws ny-raise-act and ca-sb53: this is a still-pending Massachusetts frontier-AI safety framework with Attorney General civil-enforcement authority, embedded inside a broader economic-development appropriations package rather than a standalone AI bill.