Sam Altman has put the regulatory divide between OpenAI and Anthropic squarely on the table. On October 4, 2026, Reuters reported that OpenAI CEO Sam Altman, in an interview with Politico's technology newsletter Decoded, said the benefits of AI justify society accepting some risks, and that the technology should remain broadly accessible to the public rather than locked up by a few. Asked about the difference with Anthropic on regulation, he said there is still "a lot of daylight" between the two companies.
The trade he refuses to make
Altman's words were blunt: "We believe that the world should accept some bad things happening for the benefits of this technology and people having the agency." He said he disagrees with, but understands, the opposing view: that the technology will become so powerful and so dangerous that a single lab in San Francisco should hold it, make sure nothing bad happens, and dole out the benefits. In his framing, that is "a completely unacceptable trade-off," and it runs counter to the lighter-touch regulatory stance OpenAI backs.
He did the math out loud, too: he would not trade broad access for a world with "no major hacks, no misuse of this technology, zero scams" and zero other bad things, because people will do "tremendously — orders of magnitude more — good stuff than bad stuff." In his framework, misuse is an inherent cost of openness, and the question is the ratio between cost and benefit, not whether the cost can be driven to zero.
The gap is narrowing — here is the line that remains
Notably, the remarks do not mark a return to OpenAI's old anti-regulation posture. Over the past few months the practical distance between the two companies has been shrinking: OpenAI has come out in favor of state laws imposing tougher safety requirements, its lobbyists have endorsed a bipartisan House proposal that would require top AI companies to embed outside safety evaluators, and Altman himself publicly agreed with Anthropic CEO Dario Amodei's September essay calling on the industry to slow the development of the most advanced systems to "pace the frontier." On September 23, at a UN Security Council briefing, Altman and Amodei appeared in the same forum asking for limits on their own technology. Reuters notes that high-profile hacking incidents and rogue behavior by AI models helped push OpenAI in that direction.
That makes the remaining disagreement easier to see: both sides accept that catastrophic risk needs guardrails — Altman explicitly carved out an exception for risks involving "a serious loss of control to AI" — and what they cannot agree on is the everyday level of openness. His target is always the same: using safety as the reason to concentrate access to powerful AI in the hands of a small number of companies or institutions.
Why this interview is worth reading on its own
First, it draws a boundary around the "the two labs are converging" narrative that has built since September: the consensus covers only the most dangerous tier of risk, not the more everyday question of who gets to use these systems and how much — the question that shapes the commercial landscape. Washington's political winds are blowing the permissive way as well: Treasury Secretary Scott Bessent recently criticized AI doom warnings that offer no solutions, and Altman's remarks line up with that official tone.
Second, the divide will eventually land in concrete rules: the compliance thresholds in state safety laws, how far the House's independent-evaluation proposal gets, and the cost gap between open and closed strategies will all be pulled by the public positions of these two companies. For enterprises building on either lab's models, what matters next is not whose slogan is louder, but where evaluation requirements and access policies actually change — in a statute, or in a product update.
