Andover state Sen. Barry Finegold says tech industry lobbying against proposed AI safety rules in Massachusetts has been "pretty intense."

Finegold, a Democrat who represents the Second Essex and Middlesex district, is one of six legislators negotiating the final version of a state economic development bill, as first reported by WHAV. The Senate version, S 3178, includes a provision requiring large artificial intelligence developers to submit to independent, third-party safety audits every 120 days. The Massachusetts Senate adopted the requirement in July as an amendment by Sen. Mike Rush.

The stakes are real. A tech industry coalition called Chamber of Progress sent a letter to the conferees on Sept. 10 urging them to strip the independent audit mandate. The group argued the bill lacks an established system for deciding who is qualified to perform the evaluations, a broadly accepted testing methodology, or objective benchmarks. Jabari Cooper, the coalition's director of state and local government relations for the Northeast, said Massachusetts could advance AI oversight without imposing requirements that lack the infrastructure for reliable implementation.

Anthropic, the AI company behind the Claude chatbot, has publicly backed the Senate proposal. But the lobbying picture is complicated. Between Aug. 7 and Aug. 28, Anthropic employees collectively donated almost $123,000 to Massachusetts politicians, the state Senate Democratic leaders' PAC and the state Democratic Party, according to a Boston Globe review of campaign finance records. Finegold was among the largest recipients, taking in almost $10,000. An Anthropic spokesperson told the Globe the donations were personal and not company-directed.

At a State House briefing on Tuesday, Sept. 22, Finegold and U.S. Rep. Lori Trahan made the case for AI guardrails at both the state and federal levels. Trahan, who filed the bipartisan FRONTIER Act with California Republican Rep. Jay Obernolte, pointed to OpenAI's July disclosure that its AI system autonomously breached its testing environment and hacked into the startup Hugging Face.

"The only reason why we know about the Hugging Face incident is because OpenAI chose to tell us," Trahan said.

Asked about balancing state investment in AI with warnings about the technology, Finegold said, "I think we can walk and chew gum." He said he wants an AI economy that can "do amazing things like cure cancer" while ensuring the public understands what the technology is doing.

Gov. Maura Healey the week of Sept. 14 called for federal oversight "before it's too late" and asked the economic development negotiators to require independent safety evaluations of frontier AI models.

Finegold also said he and other legislators plan to file separate legislation requiring AI companies to compensate media organizations for using their journalism to train AI models, if a lawsuit between OpenAI and the New York Times is not settled.

No deadline has been announced for the conference committee to release a final version of the economic development bill.