Altman tells Politico world must accept AI's 'bad things'

Sam Altman told Politico's Decoded podcast that "the world should accept some bad things happening for the benefits of this technology and people having the agency," using an interview with reporter Brendan Bordelon to put public distance between OpenAI and Anthropic on how far governments should go to restrain frontier AI.

He drew the line short of worst-case harms. Altman said he does not accept "the really catastrophic risks," pointing to a "serious loss of control to AI" as the category outside the trade-off.

On the rivalry, he said there is "a lot of daylight" between OpenAI and Anthropic, run by Dario Amodei, who has pushed for tighter rules on frontier models. Altman called the premise that any single San Francisco lab should steer the technology "a completely unacceptable trade-off."

The industry's political check-writing drew unusually blunt self-criticism. "Something's clearly not working," Altman said of the money AI firms have poured into the midterm cycle. "I don't think spending money on politics is what's going to make people feel better about it. The industry needs to figure out a way to address what the actual concerns are, and right now, a lot of people in the industry are being quite tone deaf." He added that "I get why AI has a really negative perception right now," and said he personally no longer backs individual candidates because AI is now so central and OpenAI does business with the government.

Altman also floated a new "liability framework" for companies whose models are capable of autonomous cyberattacks, saying "if something goes wrong with our models during training, there's going to be some version of that we need to be responsible for."

The interview was moving through the analyst-and-operator feeds we track inside a day; a dozen of the people on our radar shared it.