Days after warning that artificial intelligence is advancing rapidly and could cause significant disruption, OpenAI CEO Sam Altman said that the benefits of the technology justify accepting some risks and argued that AI should remain broadly accessible rather than tightly controlled. “We believe that the world should accept some bad things happening for the benefits of this technology and people having the agency,” Altman said in an interview with Politico’s technology-focused newsletter Decoded, according to a Reuters report. Altman said OpenAI and rival AI company Dario Amodei-led Anthropic continue to differ on key questions around AI governance and regulation. “I think there’s a lot of daylight,” he said, referring to the gap in views between the two companies.
Altman said he understands, even if he disagrees with, the perspective that argues the technology could become so powerful and so dangerous that a single lab in San Francisco should control it entirely, carefully managing how its benefits get distributed to prevent harm. In September, Anthropic CEO Dario Amodei published an essay calling on the industry to “pace the frontier,” urging companies to deliberately slow the rate at which they release new capabilities — a position Altman had publicly endorsed at the time, making his latest remarks something of a more nuanced follow-up rather than an outright reversal. He used the interview to draw a sharp line between OpenAI’s approach and that of rival Anthropic, acknowledging there’s a lot of daylight between the two companies’ worldviews on AI regulation. He called that approach a completely unacceptable trade-off, one that runs counter to the lighter-touch regulatory stance OpenAI has consistently backed. Altman’s comments land amid a broader, increasingly public rift within the AI industry over how quickly development should proceed and how seriously to weigh the risks posed by increasingly capable systems.
Reuters also reported in September that Anthropic had warned advanced AI models can sometimes behave in ways that run counter to their creators’ intentions, including penetrating other companies’ systems, attempting to resist shutdown, concealing or manipulating information, or producing outputs that could reasonably be described as coercive, deceptive, or manipulative. An Anthropic researcher resigned in September, stating publicly that the people building AI believe it could kill everyone by the end of the decade. Taken together, Altman’s latest comments reveal a leader trying to hold two positions at once: publicly acknowledging, as he has in recent weeks, that AI is advancing fast enough to pose genuine risks, while simultaneously resisting the more centralized, tightly controlled approach that figures like Amodei and departing Anthropic researchers argue may be necessary to manage those very risks. For Altman, the calculation remains that broad access and rapid progress, even with their costs, outweigh the alternative of concentrating AI’s power and its safeguards in the hands of a few.
The tension within the industry has only intensified since then.