Massachusetts Senate Advances AI Safety Bill with Catastrophic Risk Definitions Inside $325M Economic Development Package
Summary
- • Massachusetts Senate advances AI safety provisions inside $325M economic development bill; $75M earmarked to support AI development.
- • Bill defines 'catastrophic risk' as AI causing 50+ deaths or $1B+ in damages in a single incident, covering CBRN weapons, cyberattacks, and AI 'evading control.'
- • Large AI developers must build risk mitigation frameworks; Attorney General empowered to sue violators; a new AI risk commission is created.
- • 75% of Massachusetts voters back AI regulation per MassINC poll; 78% fear under-regulation more than stifling innovation.
Details
Catastrophic risk threshold defined
Bill defines 'catastrophic risk' as AI posing a foreseeable and material risk of causing 50+ deaths or $1 billion+ in damages in a single incident — one of the most specific existential risk thresholds in any US state AI law.
Covered risk categories
Catastrophic risk covers: CBRN weapon creation or release, cyberattacks, murder, assault, extortion, theft — and notably, AI that 'evades the control' of its developer or user.
Developer compliance + AG enforcement
Large AI developers must identify catastrophic risks their models pose and build mitigation frameworks. The Attorney General is authorized to sue violators, creating a civil enforcement mechanism.
New AI risk study commission
A commission led by the AG or a designee will study AI's potential risks and evaluate whether developer safety protocols are adequate — providing an oversight body to accompany enforcement.
Embedded in $325M economic development bill
AI safety provisions are part of a broader $325M economic development bill that also allocates $75M specifically to support AI development and application, defusing 'anti-innovation' criticism.
75% voter support for AI regulation
MassINC Polling Group statewide survey finds 3 in 4 Massachusetts voters support AI regulations currently under Senate consideration — strong political cover to legislate.
78% fear under-regulation
More Massachusetts voters worry about regulators moving too slowly on AI than about regulations stifling innovation, directly contradicting the industry's standard lobbying argument.
65% worried about AI harms
A majority of Massachusetts voters are more worried about AI's potential harms — particularly job losses and economic disruption — than excited about its benefits.
Targets Anthropic, Meta, Google, OpenAI
Transparency requirements would apply to Anthropic, Meta, Google, and OpenAI specifically. State also has separate House-passed bills regulating AI in political advertisements.
CBRN = chemical, biological, radiological, or nuclear weapons; AG = Massachusetts Attorney General; MassINC = Massachusetts-based polling organization; catastrophic risk = bill's legal threshold requiring mitigation frameworks; RCM = revenue cycle management
What This Means
The Massachusetts Senate's AI bill is notable for two reasons: the inclusion of AI "evading control" as a covered catastrophic risk (a rare acknowledgment of AI alignment concerns in state legislation), and its political packaging. By embedding AI safety requirements inside a $325M economic development bill that also allocates $75M to support AI growth, the legislation preemptively defuses the standard "anti-innovation" industry counter-argument. The specific catastrophic risk threshold — 50+ deaths or $1B in damages, covering CBRN weapons, cyberattacks, violent crime, and loss of AI control — is one of the most precise existential risk frameworks in any US state AI bill to date. With 75% public support, an AG empowered to sue violators, and a mandatory risk-mitigation framework for frontier AI companies, this represents a meaningful compliance burden if enacted — and a potential template for other states as federal AI legislation remains stalled.