AI safety needs a public voice
A new AP-NORC poll puts human control at the center of public AI priorities. A global health assessment adds another reason to involve the people affected.
AI safety needs a public voice. A new survey offers evidence about what Americans want from AI policy. An AP-NORC poll released October 8 found that 64% of U.S. adults think AI is developing too quickly. Another 27% judge the pace about right, while 8% consider it too slow. The concern spans party lines, though its strength differs.
The original questionnaire and results give a more useful picture than a single headline. Keeping AI under human control was an extremely or very important government priority for 81% of respondents; protecting U.S. workers scored 77%. Promoting technological innovation scored 51%. People can value useful technology while wanting stronger control over its consequences.
These are views of U.S. adults, not a worldwide verdict or a measurement of technical risk. The interviews took place September 24–28, before several incidents covered in recent briefs. The methodology describes a probability-based sample of 2,140 adults, interviewed online and by telephone, with an overall sampling margin of ±2.9 percentage points. Broad questions about pace and priorities cannot establish support for a particular law, moratorium or petition.
Today's World Health Summit launch of a Lancet Commission assessment adds a global perspective. Its agenda includes malicious AI use alongside pandemics, conflict and other threats to health. The framework compares plausible better and worse outcomes through 2100 to identify opportunities for policy action. Those scenarios are not a forecast that catastrophe must happen, and a launch description cannot substitute for examining the full assessment.
Meanwhile, AP's October 10 reporting describes industry enthusiasm for a voluntary safety agreement that calls for internal protocols and monitoring teams. Important implementation details remain unspecified.
Our view: public concern should lead to decisions people can inspect and influence. We want independent evaluations, timely incident disclosure, clear responsibility for harms, and meaningful public oversight. A company promise needs evidence of what changed and who can challenge it.
That is why we support the Statement on Superintelligence, which calls for prohibiting its development until there is broad scientific agreement on safety and controllability, alongside strong public buy-in. That is our editorial position; the poll does not prove respondents endorse the statement. Our Take action page explains the proposal and offers ways to participate. Read the wording, examine the evidence and decide for yourself.