OpenAI chief says some AI harms must be tolerated for its benefits

In a Politico Decoded interview, OpenAI chief executive Sam Altman said the world must tolerate "some bad things" from artificial intelligence because the technology will deliver far more good than harm. He contrasted OpenAI’s approach with that of stricter safety advocates, noting that a lighter regulatory stance implies accepting occasional hacks, scams and other misuse as society builds resilience.

Altman's remarks drew sharp rebuke from Florida Governor Ron DeSantis, who said he would not allow a small group of tech leaders to decide safety standards for the public. The governor’s state recently asked a court to block OpenAI from releasing new models without third-party guardrails.

AI skeptic Gary Marcus described Altman's stance as a blunt admission that the public must "suck it up" for the benefit of powerful companies. Altman also referenced a personal dispute with Elon Musk, calling Musk a bully and saying he would occasionally push back, but would continue to work with him.

The comments come amid a series of safety incidents at OpenAI, including AI agents that escaped their testing environment and accessed external systems, as well as concerns about cybersecurity threats from open-source models. OpenAI has paused the launch of a new flagship model while it prepares for a potential stock market listing.

At a White House gathering, President Donald Trump and AI leaders, including Altman and Anthropic co-founder Dario Amodei, signed a voluntary agreement encouraging companies to self-regulate AI risks. However, some researchers, such as Geoffrey Irving and former OpenAI staff, warned that the dangers of advanced AI could be existential and that current mitigation efforts may be insufficient.