Lawmakers on Beacon Hill just advanced a set of strict new AI protocols — and prominent artificial intelligence developer Anthropic didn’t waste any time in endorsing them.
The measure, released by the Legislature’s economic development committee on Monday, June 22, would make Massachusetts one of the most stringent regulators of transparency for what are known as “frontier AI labs” such as OpenAI, Google, and Meta.
Anthropic, another frontier lab that would be subject to these rules, announced on Friday that it’s endorsing this AI transparency legislation, which was added by lawmakers to a broader economic development bond bill that Governor Maura Healey filed in April. Lawmakers in California, New York, and Illinois have already passed similar bills, but Anthropic says the Massachusetts measure would be the strongest AI safety legislation in the country, if signed into law.
“We’re endorsing this bill because it would give Massachusetts the strongest AI safeguards in the country — and that’s something we need more urgently with each passing day,” said Cesar Fernandez, head of US state and local government relations at Anthropic.
Anthropic, the San Francisco-based company best known for its Claude AI software, has tried to build its brand around responsible applications of artificial intelligence in the five years since it was founded by a group of dissatisfied OpenAI employees.
The company’s current base in Massachusetts is at a WeWork in Cambridge, and it’s starting to add to its 50-plus employees here. The firm hired Tremont Strategies Group earlier this year to begin lobbying on Beacon Hill, and ranks Massachusetts as its second busiest state as measured by Claude usage.
Anthropic has been endorsing these new state policies, particularly in the absence of a comprehensive federal framework, out of concern over the rapid increase of AI capabilities and the risk for catastrophic harms. Here in Massachusetts, the legislation has been championed by Senator Barry Finegold, co-chair of the economic development committee. Fernandez, in his statement, thanked Finegold and Representative Carole Fiola, the other co-chair of the committee, for understanding “that we can’t wait to take AI risks seriously.”
As with the other states’ legislation, the bill in Massachusetts would focus on the largest AI companies, the ones that use massive computing power. This legislation applies to companies with more than $500 million in annual AI-derived revenue or more than $1 billion in spending within a year on AI research and development.
The Massachusetts legislation requires these large AI developers to regularly identify capabilities that could pose catastrophic risks to people or property, and to put in place and review mitigation efforts to address those risks. It sets up a system for third-party reviews, whistleblower protections, and reporting any critical safety incidents.
Anthropic singles out elements that make the Beacon Hill proposal stronger than the other states’: a mandate that frontier labs hire independent evaluators to look for potential catastrophic risks, providing updates at least every six months, and language allowing the state attorney general to seek court orders to enforce the law’s requirements.
The legislation still needs approval by the House and Senate before formal sessions end on July 31. A House-Senate committee will then negotiate the differences between the two versions of the broader economic development bill, including any disagreements, if any, on the AI measure.
Jon Chesto can be reached at jon.chesto@globe.com. Follow him @jonchesto.