AI’s Terms and Conditions Have a Voter Problem

The silicon salesmen have encountered an awkward customer, droogies: somebody allowed to vote.

Reuters reports that 84% of registered voters in a new Reuters/Ipsos poll see AI as a threat to American workers. That measures public opinion; it does not count jobs lost to AI.^[1]

The permissions check

• Stricter government AI regulation: supported by 80% of Democratic and 61% of Republican voters.
• Limits on building data centers: supported by 56% of registered voters.
• AI determining military-strike targets: 57% of registered voters say the U.S. government should not use it for that.^[1]

On Washington’s performance, 57% said the Trump administration had not taken AI risks seriously enough or at all; 54% said the same of Congress.^[1]

The warranty department

Reuters says six leading AI developers agreed to voluntary safety principles, including independent auditing. The agreement specifies no consequences for noncompliance, according to the report.^[1]

A lovely little warranty. Please retain your receipt and hope the promise holds.

Read the polling fine print

Ipsos surveyed 4,506 U.S. adults online through its probability-based KnowledgePanel from September 30 to October 5, 2026. The sample included 3,526 registered voters. Its published topline gives that voter sample a margin of sampling error of ±1.7 percentage points at 95% confidence; subgroup uncertainty is larger.^[1][2][3]

The available Ipsos topline omits the AI questions. The figures above therefore follow Reuters’ reporting; the exact AI question wording and response categories could not be independently checked. These results neither establish a vote to ban AI nor predict an election outcome.^[1][3]

A public asked to live with the consequences seems rather keen to see some rules. How terribly inconvenient for the sales deck.