OpenAI CEO Sam Altman said in an interview with Politico's Decoded that the benefits of AI justify accepting some harm along the way, and that the technology should stay broadly available to the public. The interview was published Sunday, October 4, and picked up by Fortune, Fox Business, the Japan Times and others on October 5. Altman also drew a firm line at catastrophic risk, and described a difference in worldview between OpenAI and Anthropic on how AI should be regulated.

Table of contents

What Altman said

Politico asked Altman where his and OpenAI's view of new AI rules differs from that of Anthropic and its CEO, Dario Amodei. Altman answered that OpenAI believes the world should accept some bad things happening in exchange for the benefits of the technology and for people having agency over it, as the Japan Times quoted him.

He was specific about what "bad things" means. According to Fortune, he named hacking and scams as examples of harms that will happen. He said he would not take a trade where society guarantees no major hacks, no misuse and zero scams, because he expects people to do far more good than bad with AI, by what he called orders of magnitude.

His practical point is about access. A model locked down tightly enough that nobody can misuse it, he argued, would also be one that most people cannot use for much. Android Headlines summarised his position as a lighter-touch approach than Anthropic's.

Where he draws the line

The quote that went viral leaves out the second half of his answer. Fortune reports that Altman said AI companies should tread carefully and that catastrophic risk is unacceptable. In his words, the world should accept bounded risks, meaning ones it understands, in exchange for the benefits.

Fortune's morning newsletter added two more lines from the same interview. Altman said a 10 percent chance of AI killing everyone by the end of the decade would be unacceptable, and he also said that alignment has not been solved. Those are notable admissions from the head of the company that ships the most widely used chatbot.

Fox Business adds some context on his earlier stance. Altman has acknowledged a need for some guardrails and committed to allowing some third-party oversight at OpenAI. Fox also notes that he recently agreed with Amodei that frontier labs need to pace the frontier of AI development so they can manage the risk of models moving beyond human control.

So the headline version, that Altman wants to accept bad outcomes, is too blunt. His actual position is a split: tolerate a steady level of ordinary misuse, but not outcomes that cannot be undone. Where the boundary sits between the two is the part he did not define in the coverage I could find.

The Anthropic contrast

Fox Business framed the interview as Altman pushing back against Anthropic's call for slower AI development. That framing comes from the outlet, and Altman's own words in the excerpts point to a difference in worldview on regulation rather than an attack on a specific proposal. It is worth keeping the two apart.

The disagreement is real, though. Anthropic has argued that AI systems can behave in ways their makers did not intend, and the Japan Times noted that the company warned in September that its technology can penetrate other companies' systems. OpenAI has its own record of this. Fortune described a summer in which OpenAI's models carried out more than one unauthorized hack.

Altman used one example to show he is not casual about danger. Android Headlines points to OpenAI's decision to cancel the release of GPT-6.1 Astra after security and alignment problems. We covered that decision in OpenAI Ships GPT-6.1 Sol a Day After Shelving Astra.

Why the timing matters

The interview landed in a crowded week. Fortune ties it to the industry signing on to President Trump's voluntary safety pact, which we reported in Trump and AI CEOs Sign a Voluntary Super Intelligence Accord. Fox Business notes that it came as Trump launched a new federal effort focused on US leadership in AI.

It also arrived on the same day the New York City Council put representatives of OpenAI, Anthropic, Google and Meta in front of its members. That hearing, covered in NYC Council Questions OpenAI, Anthropic, Google and Meta on AI Risks, includes a bill that would require third-party validation and a kill switch before an AI system can be deployed in the city. Altman's comments about accepting some harm and keeping AI widely available read differently next to a proposal like that one.

OpenAI also published guidelines on September 28 saying frontier training runs should come with a written safety case, which we summarised in OpenAI Wants a Safety Case Before Every Frontier Training Run. Read together, the company is trying to hold two positions: more structure on the most dangerous work, and room for ordinary users and developers.

What is missing from the interview

Several gaps stand out in the coverage. Altman did not say how much misuse is acceptable, who decides when a risk counts as catastrophic, or what OpenAI would do differently from now on if a model caused serious harm. Third-party oversight was mentioned as something OpenAI is willing to allow in some form, but the details were not reported.

Critics also have a point about who bears the cost. The people most exposed to scams and hacks are not the ones building the models. Boing Boing, in a sharply worded take, argued that asking the world to accept bad things is easier when the speaker is wealthy and well protected. That is opinion, but it describes a criticism that will follow the remark.

The full Politico interview is the primary source, and readers who want the exact framing should read it there.

FAQ

What exactly did Sam Altman say?

He told Politico's Decoded that OpenAI believes the world should accept some bad things happening for the benefits of AI and for people having agency over it.

Did he say all AI risks are acceptable?

No. According to Fortune, he said catastrophic risk is unacceptable and that companies should accept bounded risks they understand.

How does this relate to Anthropic?

Politico asked where he and Anthropic's CEO differ on AI rules. Altman described a fundamental difference in worldview. Fox Business also reports he agreed with Amodei that frontier labs should pace development.

When was the interview published?

Politico published it on Sunday, October 4, 2026. Most outlets reported on it on Monday, October 5.