Who decides which AI risks we accept?
Altman argues for broad access as Washington names its AI task force. Agreement on benefits still leaves hard questions about accountability.
The argument over AI safety is increasingly about who gets to decide which risks other people should bear. Two developments on Sunday, October 4 put that question in focus: Sam Altman's case for broad access to AI, and the announcement of the officials who will coordinate the White House's next steps.
Access, risk and the limits of agreement
In an interview with Politico's Decoded, reported by Reuters, Altman argued that AI's benefits justify accepting some harm rather than restricting access to a single laboratory. He described a substantial difference between OpenAI and Anthropic on regulation and favored a lighter approach. These are his arguments about the trade-off, not evidence that every avoidable harm is necessary to deliver useful AI.
That disagreement should not be reduced to a claim that Altman endorses catastrophic outcomes. The Center for AI Safety's public statement lists Altman alongside Dario Amodei and Geoffrey Hinton in supporting extinction-risk mitigation as a global priority. The statement does not specify a pause or a regulatory mechanism. Shared concern leaves considerable room for disagreement about what to require.
Washington names its team
AP reported that President Trump named national intelligence director Jay Clayton to lead the federal SuperIntelligenceForce. The group will coordinate government efforts and consult consumers, public-interest groups, religious organizations, infrastructure providers and AI companies. Other members include FTC chair Andrew Ferguson, Pentagon technology chief Emil Michael and personnel director Scott Kupor.
A separate Reuters report citing the Wall Street Journal says Clayton's team is expected to deliver recommendations within 120 days, examining AI's opportunities, risks and the federal government's oversight role. A timetable for recommendations is not itself an operating safeguard; the announcement does not establish how the eventual proposals will be enforced.
Our view: make the decisions inspectable
For readers, the useful questions are specific. What evidence will outside evaluators see? Will affected people have a meaningful voice? Which findings would require a developer to change course, and who can make that happen?
We can value access to ChatGPT while asking for independent scrutiny and enforceable protections. Broad benefits do not settle how burdens should be distributed. Watch for published recommendations, clear responsibilities and ways to challenge decisions—not just agreement that AI should be beneficial. Our Safety & Society guide separates practical precautions, evidence, policy arguments and frontier research so those claims can be examined.