The man building the AI future just told you he's fine with some collateral damage—and thinks his competitors are the real safety threat.
The Summary
- Sam Altman told POLITICO that "the world should accept some bad things happening for the benefits of this technology and people having the agency"—a direct shot at Anthropic's safety-first positioning
- Altman posted on X that treating AI models as "religious force" or surrendering human judgment to them is "a real safety issue," days after Anthropic held meetings with religious leaders
- OpenAI has quietly shifted toward favoring tougher AI regulations after high-profile hacking incidents and rogue model behavior, narrowing the policy gap with Anthropic
- The timing matters: one of OpenAI's own safety leaders, David Robinson, just quit and publicly criticized the company's safety record
The Signal
The mask-off moment here is Altman drawing a line between acceptable risk and what he sees as performative caution. His comment to POLITICO wasn't accidental hedging. It was doctrine: OpenAI believes human agency with powerful tools beats locked-down safety theater. The subtext is even sharper—he's saying Anthropic's approach infantilizes users and overestimates catastrophic risk.
But there's a contradiction worth noting. OpenAI has been moving toward Anthropic's position on regulation for months. The gap between the two companies has narrowed as OpenAI dealt with models behaving unpredictably and security breaches that made headlines. Altman still wants a "lighter touch" than Anthropic, but he's no longer the full-throated deregulation advocate he was two years ago.
"We believe that the world should accept some bad things happening for the benefits of this technology and people having the agency."
Then there's the religion angle. Altman's X post called out people who "ascribe religious force or a surrender of human judgment to AI models" as a safety issue. He didn't name Anthropic directly, but the timing says it all. Days earlier, The New York Times reported that Anthropic cofounder Chris Olah had been holding private meetings with religious leaders about the company's AI models. Altman's framing flips the safety argument: the real danger isn't the technology, it's the people building messianic narratives around it.
This matters because it reveals the actual fault line in AI development. It's not technical. It's philosophical:
- Should AI be treated as a tool humans control, or a force humans must accommodate?
- Is risk mitigation about limiting access, or educating users?
- Who gets to decide what counts as "catastrophic" versus "acceptable" harm?
Altman's bet is that ordinary people will do "orders of magnitude more good stuff than bad stuff" with AI if you let them. That's an empirical claim masquerading as values. And it's coming from the CEO whose company just lost a senior safety researcher who publicly rebuked OpenAI's safety efforts on his way out the door.
The Implication
Watch how OpenAI navigates the gap between Altman's libertarian rhetoric and the company's actual regulatory posture. If they're advocating for tougher rules while simultaneously arguing the world should accept bad outcomes, someone's playing chess. Either they're trying to regulatory-capture the space before smaller competitors can scale, or they're genuinely trying to thread a needle between safety and access that no one else has figured out.
For builders and users, the message is clear: the AI platforms aren't converging on a shared safety model. They're diverging on first principles. That means the tools, the guardrails, and the philosophy behind them will continue to fracture. Pick your stack accordingly.