A new MassINC Polling Group survey of 1,264 registered voters, conducted June 23 to July 1 and sponsored by AI giant Anthropic, found that 65% say they're more concerned about AI's dangers than excited about its potential to aid humanity.
Who Pays for the Machine
Anthropic paid for the poll. That matters. The company sits at the center of the very system people say they don't trust, and the survey it sponsored still found a public that sees the risks more clearly than the sales pitch. Only 16% say they're excited about AI's potential to help humanity, while 65% say they're more concerned about its dangers. The numbers come from 1,264 registered voters in Massachusetts, surveyed over the same year, and they point to a basic split between corporate promise and lived suspicion.
Among the nearly two-thirds with concerns about AI, job loss and economic disruption top the list of worries at 32%. Loss of skills or overreliance on AI and the spread of misinformation each rank at 16%. Environmental impact from data centers and misuse of AI power each rank at 15%. Those are not abstract fears. They're the costs of a system built to concentrate power, automate labor, and push the fallout onto everyone else.
What People Actually Said
Usage is split roughly down the middle. Thirty-nine percent use AI tools like Claude, ChatGPT, Gemini or Microsoft Copilot every day or a few times a week, while 27% say they haven't used one at all in the past year. Even among those who don't use AI, 80% say they've heard "a great deal" or "a fair amount" about AI, and 72% say they're at least somewhat familiar with generative AI tools. The apparatus is everywhere, whether people opt in or not.
Voters also back government oversight of AI. Fifty-nine percent say the government should oversee AI companies that deploy models with serious risks, while only 23% trust companies to police themselves. That gap says plenty about who people think should not be left in charge. Seventy-eight percent in Massachusetts worry more about the dangers of failing to regulate AI than about rules slowing tech innovation. Nearly half, 47%, say it's "very likely" that AI could enable catastrophic harms like chemical, biological or cyberattack risks without safeguards.
Beacon Hill and the Limits of Permission
Awareness of local efforts is lower. Just 13% have heard "a great deal" or "a fair amount" about Beacon Hill's efforts to pass AI safety rules. Once informed of the Legislature's intentions, 75% back requiring the largest AI companies to assess catastrophic risks, publish safety plans and report serious incidents to the attorney general. Two-thirds want the state to act now rather than wait for Washington to assemble federal guidelines.
That last detail shows the familiar trap. People are told to wait for lawmakers, then wait for federal guidelines, then wait for the companies to behave. Meanwhile, the firms keep building systems that voters say could bring job loss, misinformation, environmental damage and even catastrophic harms. The public sees the danger. The institutions move at the pace of their own interests.
The poll itself was conducted June 23 to July 1, and it captures a blunt contradiction: many people use the tools, but many more fear what those tools are doing to work, knowledge and safety. Anthropic may have sponsored the survey, but it couldn't sponsor away the distrust. The numbers didn't cooperate.