Three-quarters of Massachusetts voters favor regulations set to be discussed this week on Beacon Hill that would create a commission to help regulate AI companies, according to a poll released by the MassInc Polling Group on Monday.
The release of the poll comes as the Massachusetts Senate deliberates an economic development bill that calls for a commission to consider AI regulatory and safety issues, like the hiring of third-party auditors to verify that developers are following safety protocols. The bill features other AI guardrails as well, including requiring major AI developers to create safety frameworks.
The proposed economic development bill that includes AI guardrails is slated for floor debate on July 23.
As AI becomes increasingly used by many on a regular basis, 75% of state voters said they would either strongly support or somewhat support the proposed AI regulation bill.
The poll also found that at least 65% of voters across Massachusetts are concerned about the potential dangers of artificial intelligence,
The poll was sponsored by Anthropic, maker of the AI platform Claude. The San Francisco-based company has long supported stronger government oversight of the industry than many of its competitors.
The poll also finds that higher-income voters are more likely to be excited about how AI can help humanity. But even within this group, at least half said they were more concerned than excited.
“I think a lot of the risks of AI are very pervasive and sort of hidden — like the fact that your doctor’s using it, your lawyer’s using it, and the way that may interact with you,” John Basl, a professor of philosophy at Northeastern University. “I think there’s a sense in which these numbers probably underestimate the risk that people are concerned about, just because it’s focused on these big one-shot risks.”
Without regulations and safeguards, at least 82% of respondents said they feel that AI systems could create “serious risks,” like “enabling chemical, biological, radiological or nuclear harms, large-scale cyberattacks.”
But despite concerns about AI among poll respondents, the majority of people are still using it. According to the poll, 61% of respondents reported using AI tools like ChatGPT, Claude, Google Gemini or Copilot on a monthly basis. Nearly 40% reported using AI tools multiple times a week.
“I think people feel that that train is maybe going a little too fast, and we’re still dealing with the last wave of impacts from technology, and now it’s just going to keep going faster,” said Andrew Doris, policy director at Secure AI Project. “I think they [people] feel a little disoriented, they feel a little powerless, and they want to exert what power they have through their elected representatives to do something about it.”
The poll finds that respondents generally support a list of nine regulatory steps, including: reporting urgent safety problems within 24 hours; protecting AI company employees who report safety problems; updating their safety plan at least once a year; and paying up to $1 million for each violation.
While Republican voters are less supportive, average support for the nine proposed regulations across all poll respondents is 56%.
Doris said that for AI regulations to be successfully implemented, there needs to be a “transparent, lawful process that the government has to evaluate these risks and identify calibrated interventions to mitigate those risks.”
“I think right now, the government’s kind of caught flat-footed by some of this and reaching for tools that are not fit for purpose, and they lack that actual system in place to know how to handle this,” he said.
Doris added that the process also needs to be informed by technical experts who are both independent from AI companies and political pressures.
“Ultimately, although this isn’t in any of the state-level bills at the moment, I think we need to be able to empower a government official somewhere to review these judgments from these experts and ultimately intervene to block unsafe behavior,” he said.
The majority of voters said they would prefer to see Massachusetts lead the way on setting AI safety standards instead of waiting for other states or the federal government to act.
Beauclarine Thomas, state and local government relations lead at Anthropic, said California and New York advanced robust AI safety laws last year, and just last week, Illinois signed “the strongest AI safety bill” into law.
“States are absolutely in the driver’s seat, and the question isn’t between state action versus federal action, it’s a question of state action versus no action and letting the technology evolve without safeguards set in place,” she said. “With states, they’re tasked with protecting roads, protecting infrastructure and that’s some of the things that the bill would protect.”
