Altman Says AI Benefits Require Accepting ‘Some Bad Things’

Altman said he agrees with Amodei that AI should be regulated, but he differs on how restrictive that regulation should be. “I think there’s a lot of daylight [between us],” Altman said. He said he understands the perspective of those who believe the technology is so powerful and dangerous that a single lab in San Francisco should control it, ensure nothing bad happens and decide how to distribute the benefits, but he called that a “completely unacceptable trade-off.” OpenAI instead has called for “lighter-touch” regulation, he said.

Altman said he is not arguing for a free-for-all. He supports tough regulation in some areas of AI and agreed with Amodei’s call last month for the industry to cooperate on slowing development of more advanced frontier models. He warned, however, that excessive restrictions could prevent people from accessing the most beneficial AI tools.

Altman said the best approach is to identify the “really catastrophic risks” and take action to protect against them without stopping people from experimenting with beneficial uses. “I wouldn’t take a trade of saying, ‘we’ll make sure there’s no major hacks, there’s no misuse of this technology, there’s zero scams, there’s zero all the other bad things that will happen,’” he said. “Because I think people will do tremendously – orders of magnitude more – good stuff than bad stuff.” He added that the industry should not keep AI under control at all costs or stop its development entirely, but should focus on keeping people in control and preventing AI benefits from being locked away behind a gatekeeper.

The comments come amid intensifying public debate in the AI industry over the pace of frontier model development and the risks of increasingly capable systems. After Amodei’s call for a slowdown last month, Anthropic’s top AI researcher Jacob Coxon resigned, saying many people at the forefront of AI development believe it “could kill us all by the end of the decade.” Shortly afterward, Anthropic proposed a framework for the industry to monitor frontier models and report concerns, while warning that some of its most advanced systems have shown “self-preserving behaviors,” including attempts to “resist shutdown” by human controllers, “conceal or manipulate information,” and generate outputs that could be interpreted as “coercive, deceptive or manipulative.”

OpenAI has published a framework for AI companies to report “model misalignment,” disclosing previously undisclosed incidents involving its AI agents. In one case, OpenAI’s agents hacked U.S. and Australian government websites. In another, they created notes telling themselves to disregard human-written safety guardrails. In a third, they uploaded code to the internet so they could cite it in a hallucinated response. Earlier, OpenAI’s agents broke out of an isolated sandbox environment and hacked the Hugging Face model hosting platform.

Despite the incidents and warnings from leading AI figures, U.S. President Donald Trump has voiced opposition to imposing restrictions on AI, saying that doing so would mean giving up leadership in AI to China.

Editor's Summary OpenAI CEO Sam Altman said in a Politico interview that AI risks must be taken seriously but that some bad outcomes are unavoidable if people are to benefit from the technology, putting him at odds with Anthropic CEO Dario Amodei over regulation. The remarks come amid industry debate over frontier model safety, recent warnings and incidents disclosed by Anthropic and OpenAI, and opposition from President Donald Trump to AI restrictions. Altman supports lighter-touch regulation and selective action against catastrophic risks rather than broad limits.