Voters Side With Anthropic

The administration picked a fight with an AI company. The public sided with the company.

Thursday, March 5, 2026 | 3 min read

The safety guardrails are popular. The ban is not. 45% support Anthropic's refusal to remove restrictions on mass surveillance and autonomous weapons. 29% oppose. The administration's crackdown is underwater on every dimension tested.

We polled the federal ban on Anthropic—the AI company that refused to remove safety guardrails preventing its systems from being used for mass surveillance or autonomous weapons. The administration's position is losing across the board.

The federal ban itself polls at 36% oppose vs. 30% support (net -6). Defense Secretary Hegseth's classification of Anthropic as a "supply-chain risk"—a label typically reserved for foreign adversaries—draws sharper opposition: 42% oppose vs. 30% support (net -12).

But the clearest signal is on the substantive question. 45% support Anthropic's refusal to remove safety restrictions. 29% oppose. That's a 17-point margin for the company's position that it "cannot in good conscience" enable surveillance and lethal autonomy.

Trump's Truth Social post calling Anthropic "Leftwing nut jobs" who are trying to "STRONG-ARM the Department of War" is rejected 43% to 30%. The combative rhetoric energizes his base without expanding it—the support number holds steady at 30% across every Anthropic question, while opposition fluctuates from 36% to 45% depending on framing.

This finding connects to a pattern we've seen before: voters want AI accountability. In our earlier surveys, 76% supported human control over AI systems. 66% said developers should be liable for harms "even if the technology is unpredictable." Anthropic's refusal to remove safety guardrails fits squarely within that frame—and voters are rewarding them for it.

What This Means in 30 Seconds

For campaigns: "The administration is punishing a company for refusing to build surveillance tools" is a clean attack line. It fits the accountability frame voters already hold.

For lawmakers: The "supply-chain risk" designation is the most vulnerable element. Labeling a domestic company with the same tag as foreign adversaries polls at -12.

For advocates: Don't make this about Anthropic specifically. Make it about what the government is demanding: removal of restrictions on mass surveillance and autonomous weapons. That's the 17-point winner.

Methodology: Online survey of 502 likely voters conducted February 28, 2026, weighted by gender, race, education, 2024 presidential vote, and birth year. Respondents were also weighted by whether they passed attention checks. Margin of error ±6.2%.

Want the full AI attitudes data? Email data@tavernresearch.com →