BU’s AI leaders weigh in on Healey’s AI regulations as lawmakers debate bill

A Massachusetts economic development bill, which could set precedent for artificial intelligence regulation, sits before a legislative committee for reconciliation after differing versions were passed by the state House and Senate.

The Center for Computing and Data Sciences at Boston University. A bill proposed by Gov. Healey would require major AI models to get evaluated by third parties at least three times a year. (EMERSON LIHN/ DFP PHOTOGRAPHER)

In September, Gov. Maura Healey called on the state legislature to require independent evaluations of the most powerful AI models under the Mass Wins Act. Major AI developers, like OpenAI, Anthropic and Meta, would be required to pay independent experts to evaluate their models at least three times a year for risks of catastrophic harm.

If enacted, Massachusetts will be the first state to allow third parties to look into the technology. States like California, Illinois and New York only require AI developers to publish their own plan for managing catastrophic risks.

Junior Denali Schlesinger, president of Boston University’s AI Society, said Healey’s bill is a sign of progress but potentially has too narrow of a focus.

“This bill only targets really, really large models like Frontier, Fable, Opus 5.5,” Schlesinger said. “A lot of the danger is going to be from smaller open source models that have the capability to do cyber attacks, to become misaligned, but they just aren’t really thought about as much.”

Schlesinger said one state requiring evaluators for AI models means every state will benefit.  However, some legislation that targets risks might be misdirected, he said, as AI is dependent not only on training data, but on the developers who construct the architecture and training methods.

“I don’t think the AI itself is necessarily the problem,” Schlesinger said. “It’s more the people in AI that I’m worried about.”

Tim Duncan, a lecturer and director of the AI program at BU’s School of Law, said laws like Healey’s are flawed.

Regulatory frameworks might not effectively hold AI companies accountable, he said. Some laws may protect companies from lawsuits while creating oversight systems that do little to prevent harm, Duncan said. 

“When one of these laws is passed, it gets a lot of publicity, but then it takes away people’s rights to sue,” Duncan said. “So it protects them against their biggest potential liabilities, while creating this new administrative ambiguous regulatory regime that generally [is] ineffective.”

Regulating tech companies should be a responsibility at the national level, he said.

“Whatever states do is not going to be particularly effective,” Duncan said. “Obviously, that’s going to be tough to do with the Trump administration as it stands right now. But I think the piecemeal approach, state by state, is going to be tough.”

The Trump administration has prioritized reducing regulatory barriers for AI developers and established a voluntary framework for testing advanced AI models.

In an interview last month with MassLive, Healey acknowledged that enacting statewide policy in clash with federal action is not the most productive. However, she does not anticipate President Donald Trump will take action to pass similar regulatory measures.

State Senator Barry Finegold, chair of the legislature’s Committee on Economic Development and Emerging Technologies, championed the Mass Wins Act as it moved through the Senate.

“I sponsored this legislation because AI has the potential to create incredible prosperity,” Finegold wrote in a statement to The Daily Free Press. “But innovation without guardrails is not progress.”

In a poll by the Artificial Intelligence Policy Institute in June, 68% of respondents said they support the government formally reviewing the most advanced AI models before public release.

Last month, Anthropic CEO Dario Amodei wrote in an essay that he too endorses building AI at a “balanced rate” to ensure its safety, while still making technological advancements. 

John Byers, a computer science professor and head of BU’s AI Development Accelerator, said Amodei’s comments are proof tech companies want some regulation. 

“The big challenge is that it’s got to be a level playing field,” Byers said. “It’s very hard to work that all actors worldwide are going to comply with regulations. So this honest disclosure, the transparent disclosure, I think is the most important part of the regulation.”

Read more here: https://dailyfreepress.com/10/26/23/221826/bus-ai-leaders-weigh-in-on-healeys-ai-regulations-as-lawmakers-debate-bill/
Copyright 2026