OpenAI CEO Sam Altman says the benefits of artificial intelligence justify accepting some risks, and that the technology should remain b...
"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.
It's a clear, deliberate statement of priorities, not a side comment. And it draws a sharper line than OpenAI has usually drawn in public between its own approach to AI deployment and the more cautious posture coming out of Anthropic.
The Core Argument: Broad Access Over Centralized Control
Altman framed the alternative position directly. He said he understands, but disagrees with, the view that AI is becoming powerful and dangerous enough that a single lab in San Francisco should hold it and carefully manage how its benefits get distributed. He called that a completely unacceptable trade-off, one that runs counter to the lighter-touch regulatory stance OpenAI has backed.
The logic underneath it is a numbers bet: Altman has said he expects people to do orders of magnitude more good than bad with AI as access widens. Risks like hacking, scams, and fraud are treated as an expected cost of that access, not a reason to restrict it.
The Divide With Anthropic
Altman was explicit that there's real distance between OpenAI and Anthropic on regulation. "I think there's a lot of daylight," he said.
That distance has been building publicly for weeks. Anthropic CEO Dario Amodei published an essay in September calling on the industry to "pace the frontier," arguing for a more deliberate, slower approach to deploying increasingly capable systems. Anthropic has also separately warned that AI can act in ways that run counter to its makers' intentions, including penetrating other companies' systems, a concern that has shaped its more cautious public posture.
Altman's comments land as a counterpoint to that framing: not that the risks Anthropic describes are fake, but that treating risk minimization as the top priority is itself a dangerous concentration of control.
Why the Timing Matters
These comments don't land in a vacuum. They follow a string of AI agent security incidents across the industry, and they come as the policy conversation between leading labs has been shifting back and forth between calls for caution and calls for access. Altman has also publicly warned, separately from this interview, about two specific failure modes: humans losing control to AI systems, and a single entity using AI to concentrate power. That's worth holding next to this statement, since it shows his position isn't "no risk matters," it's a specific bet on which risk is worse: distributed harm from broad access, or concentrated control in the name of safety.
What This Means If You're Building on These Models
You don't need to take a side in a disagreement between two frontier labs' CEOs to act on this. What's useful here is the signal: the companies you build on don't share a single philosophy about how much risk is acceptable, and that's going to show up in product decisions, release pace, and default safety settings.
That's a reason to build your own guardrails rather than inherit a vendor's risk appetite by default: least-privilege access for anything agentic, independent monitoring outside the model's own self-reporting, and a documented incident response plan that doesn't assume the underlying platform will catch every failure mode for you.
If your team is deploying AI across multiple vendors or models and wants a consistent risk and safety posture that doesn't depend on any one lab's philosophy, ATX Soft can help you build that layer yourself.
Frequently Asked Questions
What did Sam Altman actually say?
Altman said the world should accept some bad outcomes from AI in exchange for the technology's benefits and people having agency over it, arguing AI should stay broadly accessible rather than tightly controlled by a single company.
Where and when did he say this?
In an interview with Politico's technology newsletter Decoded, published as a podcast on Sunday, October 4, 2026.
How does this differ from Anthropic's position?
Altman said there is real daylight between OpenAI and Anthropic on AI regulation. He characterized the alternative view, that a single lab should control powerful AI to prevent harm and carefully dole out benefits, as a completely unacceptable trade-off. Anthropic CEO Dario Amodei had separately called for the industry to pace the frontier and slow down.
What kinds of risks is Altman referring to?
Reporting points to risks like hacking, scams, fraud, and other misuse. Altman has argued people will use AI for far more good than harm, even as broader access increases the chance some of that harm occurs.
Does this mean OpenAI is ignoring AI safety concerns?
Not exactly. Altman has publicly warned about two specific failure modes: humans losing control to AI systems, and a single entity using AI to concentrate power. His comments reflect a tradeoff position, prioritizing broad access and benefit over tightly centralized control, not a dismissal of risk entirely.
