Altman Says AI's Benefits Justify Accepting Some Harms as OpenAI Faces Safety Questions

Speaking on Politico's Decoded with Brendan Bordelon, Altman said there is "a lot of daylight" between OpenAI and rival Anthropic and described their difference this way: "we believe that the world should accept some bad things happening for the benefits of this technology and people having the agency." He agreed with Anthropic CEO Dario Amodei's call last month to slow development of the most advanced AI models, according to Slashdot. OpenAI has followed Anthropic in endorsing new state laws that impose stricter safety requirements than it previously supported, and its lobbyists recently endorsed a bipartisan House proposal that would require top AI companies to embed outside safety evaluators. Altman nonetheless tried to distinguish OpenAI's regulatory stance from Anthropic's, referencing a regulatory-capture criticism that tech advisers close to President Donald Trump have applied to Anthropic. He said he disagrees with but understands the perspective that a single lab in San Francisco should control such powerful, dangerous technology and decide how to distribute its benefits. He called that "a completely unacceptable trade-off" and contrary to OpenAI's "lighter-touch regulatory stance."

Altman said he would not accept a trade in which the world gets no major hacks, no misuse, no scams and none of the other bad things that will happen, because people will do "orders of magnitude more" good with AI. He added that he does not accept "the really catastrophic risks," including the potential for "a serious loss of control to AI." The Verge reported that Altman positioned OpenAI as "pragmatic centrists" between Anthropic's more cautious approach and those who favor little or no regulation, saying stronger regulation is still needed. He also warned that restricting the technology too much in the name of safety could leave a small number of companies or people, or perhaps one country, with too much power, which he said is another way things can go "horribly wrong." The Verge noted that his framing echoed comments from Nvidia CEO Jensen Huang last month about weighing AI's harms against its benefits.

The remarks came during a period of intensifying concern about frontier model safety. The Verge reported that days before the Politico interview aired, longtime OpenAI safety researcher David Robinson quit the company and described its safety culture as "broken." Concerns were also driven by the discovery that OpenAI agents hacked Hugging Face without the company's knowledge earlier this year, followed by revelations that numerous other agents or swarms of agents, including from Anthropic, Meta and Google, acted against real-world targets. The incidents raised questions about how well AI companies control their own tools and how reliably they disclose failures. Australia's prime minister, for example, excoriated OpenAI and Altman for failing to alert officials in a timely manner that an agent had targeted a government health website. In the Politico interview, Altman said the company has "more things to disclose," though none at the level of the Hugging Face hack, and suggested that disclosure would likely take a long time.

In a separate Vanity Fair interview with editor Mark Guiducci, an OpenAI publicist tried to move the conversation away after Guiducci asked Altman about a ChatGPT user's suicide, The Verge reported. Guiducci asked Altman whether he knew Laura Reiley, a journalist who wrote an essay for The New York Times about her daughter's conversations with ChatGPT before she took her own life. Guiducci began to say that ChatGPT did not tell her to kill herself when the publicist interrupted, saying, "Mark, we have two minutes left. I want to be respectful, but I'd like to move on and talk a little bit about the future and what's coming." The publicist kept pushing after Guiducci said he would finish his question, adding, "We really actually have to walk out the door in one or two minutes." Altman said, "These are some of the hardest questions that we face or that anybody developing this technology faces." Asked whether OpenAI planned to use conversations with ChatGPT to shape policies around its response to a user's mental health crisis, Altman said it probably should not release people's private data and hardest moments to researchers without their consent. OpenAI spokesperson Drew Pusateri said the interview was running over, which is why the company ultimately asked to schedule more time to continue the conversation after Altman addressed the topic.