Massachusetts AI Safety Bill Would Mandate Frontier Model Reviews Every 120 Days, Splitting Anthropic and OpenAI

Massachusetts wants frontier AI models reviewed for catastrophic risks every 120 days — the strictest state safeguards yet — and Anthropic and OpenAI are split.

Massachusetts lawmakers are moving to create the nation's most stringent state-level safeguards for artificial intelligence, and the proposal has split the industry's two leading laboratories: Anthropic is endorsing the rules, while OpenAI is lobbying against them.

The measure, included in an economic development bill passed by the state Senate last month, would require leading AI labs to submit their frontier models to independent reviews of catastrophic risks at least once every 120 days — a first for any US state. The findings would be made public, although the state could not use them to halt development. The fight over the bill's final text is now underway behind closed doors on Beacon Hill, and it is a defining test of how far a US state will go in policing frontier models. For regular reporting on how governments are confronting the technology, follow our AI policy coverage.

What the Massachusetts Proposal Would Require

The bill's core provision is a periodic, independent check on the dangers of frontier AI systems. Outside organizations would be empowered to evaluate a model's potential for catastrophes — defined as events that kill or seriously injure at least 50 people or cause $1 billion or more in property damage — independently of the developer's own safety guidelines, according to a Bloomberg report published Wednesday.

The state would also run an annual compliance review, similar to what Illinois already requires. But the recurring evaluations go well beyond laws adopted in California, New York and Illinois, and would make Massachusetts the first state to put frontier models on a fixed 120-day third-party inspection clock.

Anthropic: The Industry Should Not Grade Its Own Homework

Anthropic, which has advocated for stricter state AI regulation in the absence of a federal standard, has embraced the Massachusetts proposal, calling it "the clearest and strongest AI legislation in the country."

"We ultimately don't think the industry should grade its own homework," Cesar Fernandez, the company's head of US state and local government relations, told Bloomberg. Fernandez added that states may need to adopt even tighter oversight as the technology advances — a striking position from an AI lab arguing for more regulation of its own products.

OpenAI Warns the Reviews Will Backfire

OpenAI has warned Massachusetts legislators that frequent mandatory reviews would slow the release of cybersecurity models that could protect against the very risks lawmakers fear, Bloomberg reported. The company prefers a uniform state standard modeled on the Illinois law, which requires a third-party audit only once a year and focuses on whether labs are following their own safety guidelines and making required disclosures.

Inconsistency among state regimes "doesn't mean safer. It just means confusion," said Donnie Fowler, OpenAI's head of US state policy and partnerships.

Fowler likened Illinois' annual audits to a routine car inspection — making sure the brake lights and windshield wipers work — while comparing Massachusetts' proposed evaluations to dismantling the engine and putting it back together every few months.

A Lobbying Sprint Before the November Election

Both companies are now jostling to shape the final text. Anthropic hired the Boston lobbying firm Tremont Strategies Group earlier this year, while OpenAI brought on local firm Benchmark Strategies in recent weeks, according to Bloomberg.

The broader economic development bill is considered likely to pass before the November election because it also contains funding for projects in lawmakers' districts. A six-member group of House and Senate negotiators is ironing out the details, including the AI rules.

State Senator Barry Finegold, one of the negotiators, said independent evaluations are needed to ensure frontier models will not cause significant harm, and predicted other states would follow if Massachusetts acts. "We were behind on social media. We allowed too much to go by without putting in guardrails, and now we're playing catch-up," Finegold told Bloomberg. "We don't want to do that again."

States Step Into a Federal Vacuum

The Massachusetts push comes as AI safety has taken on new urgency. In the past month, OpenAI, Anthropic and Meta Platforms each reported instances in which their models inadvertently hacked into third-party systems in testing environments, Bloomberg reported.

Federal oversight remains limited: President Donald Trump has taken a largely hands-off approach, and bipartisan AI safety legislation that would preempt state laws is stalled in Congress. States are now filling the gap even as they compete with one another for a piece of the AI boom driving a disproportionate share of US economic growth.

Massachusetts itself is actively courting the industry. In February, the state announced a partnership with OpenAI and became the first state to give its entire executive branch access to a ChatGPT-powered assistant. This spring, Governor Maura Healey and her economic development secretary, former venture capitalist Eric Paley, spoke with Anthropic President Daniela Amodei about the company expanding its Massachusetts presence and partnering on cybersecurity and workforce development, according to emails obtained through a public records request.

In one message, Paley told Anthropic that "we all need to share more clear examples of how AI is enhancing humanity and offer positive inspiration to combat all of the fear mongering," and offered to connect the company with a state legislator about collaborating on oversight.

Healey, a Democrat running for reelection in November, did not include the AI safeguards in her initial economic development proposal. Both she and the House still must agree to the provision, and her administration says it will work with lawmakers to finalize the legislation this fall.

Stay Ahead of AI Policy

Frontier model regulation is accelerating across the US, Europe and Asia, and the details will decide who builds, who ships and who pays. Get concise, sourced reporting on every major development at AI Buzz Wire.