The Pulse
Altman Says AI Benefits Justify Accepting Some Harm
OpenAI CEO Sam Altman said society should accept some AI-related harms in exchange for the technology’s benefits and people’s ability to use it. He distinguished those harms from catastrophic risks, including a serious loss of control to AI
AI.info Team · Written with AI ·
OpenAI chief executive Sam Altman says society should accept some AI-related harm rather than restrict access to the technology to prevent every misuse. His position highlights a continuing difference with Anthropic over how much risk regulation should tolerate, even as the companies have found some common ground on safety measures.
Altman rejects a zero-harm bargain
In an interview with POLITICO’s Decoded, Altman argued that AI’s benefits and people’s ability to use the technology justify accepting some bad outcomes. He said broad access should not depend on a promise that misuse, scams or other harms can be eliminated.
“We believe that the world should accept some bad things happening for the benefits of this technology and people having the agency,” Sam Altman, OpenAI’s chief executive, said.
Altman described his regulatory approach as “lighter-touch,” saying it accepts that some bad outcomes will occur as society develops resilience. He also rejected a bargain that would guarantee there were no major hacks, misuse, scams or other bad things before people could use the technology. He argued that people would do far more good than bad with AI.
His position did not treat every risk as tolerable. Altman said he would not accept “really catastrophic risk,” including “a serious loss of control to AI.” The interview account does not specify a detailed test for where that boundary falls or what safeguards would be sufficient to prevent such an outcome.
Different proposals, limited agreement
Altman described a difference in outlook between OpenAI and Anthropic and its chief executive, Dario Amodei. In the interview account, the disagreement concerns how far to go in limiting powerful AI and whether the technology should remain broadly accessible, rather than whether safety deserves attention.
Amodei separately called for slowing the pace at which AI models’ capabilities improve. In a September 12, 2026, public response, Altman wrote, “I agree with Dario that we need to pace the frontier.” Axios also reported that Altman said OpenAI would give external evaluators access. That public response concerned pacing and evaluation; it was separate from his later remarks in the POLITICO interview.
The distinction matters to the policy debate. Support for outside evaluation does not by itself establish a position on every proposal for oversight, just as Altman’s comments on access do not amount to a claim that all safeguards should be rejected. His stated objection is to a trade-off that, in his view, would concentrate control in one lab or make public access contingent on preventing every harmful use.
Access weighed against catastrophic risk
Altman’s remarks set out a boundary between harms he says society should accept and risks he says it should not. Scams, misuse and other bad outcomes appear on the first side of that distinction; a catastrophic outcome such as a serious loss of control to AI appears on the second. The interview account leaves unresolved how that boundary would be applied in practice.
Altman’s public support for pacing and external evaluation sits alongside his argument for broad access. In the interview, he opposed a guarantee against every bad outcome, while saying catastrophic risks were unacceptable. How those positions might work together—and whether oversight can address catastrophic risks without restricting broad access—is not answered in the reported remarks.