Sam Altman Argues Some AI Harms Are Worth Accepting to Preserve Broad Access

OpenAI CEO Sam Altman is arguing that preventing every harmful use of AI would come at too high a price: denying people its benefits and control over how they use it. In an interview with POLITICO’s Decoded published October 4, he called for accepting some harms while drawing a firmer regulatory line around catastrophic risks.

A risk threshold, not a free-for-all

Altman’s proposed approach starts by identifying the “really catastrophic risks” and protecting against them. Outside that boundary, he wants people to retain room to experiment with beneficial uses. He supports tough regulation in some areas, but favors lighter-touch rules where restrictions would limit access to increasingly capable tools.

The distinction is between guarding against the gravest dangers and trying to guarantee that nothing goes wrong. In comments quoted by SiliconANGLE, Altman rejected a trade that would eliminate major hacks, misuse, scams and other harmful outcomes at the expense of AI’s benefits. Those examples make his position more concrete than a general appeal for less regulation.

His justification is a prediction about what people will do with access, not a claim that misuse disappears. He expects beneficial activity to outweigh harmful activity by a large margin. That expected balance is the basis for his willingness to accept some bad outcomes rather than make their complete prevention the condition for access.

Because I think people will do tremendously – orders of magnitude more – good stuff than bad stuff.

Sam Altman, speaking to POLITICO’s Decoded, as quoted by SiliconANGLE

Who gets to distribute the benefits?

Asked about his differences with Anthropic CEO Dario Amodei, Altman said there was “a lot of daylight” between them. He described a position he understands but rejects: powerful AI remaining with a single San Francisco lab that ensures nothing bad happens and decides how to distribute the benefits.

He called that a “completely unacceptable trade-off.” This was Altman’s characterization of the restrictive position, not a direct statement from Amodei. His objection concerns who holds authority as well as how much harm is prevented: he wants people to remain in control, rather than have AI’s benefits distributed by a gatekeeper.

Slower development can coexist with wider access

Despite that disagreement, Altman said he agrees with Amodei’s call for the industry to cooperate on slowing development of more advanced frontier models—the most capable systems being built. His support for that restraint sits alongside, rather than replaces, his argument for broad access.

The two positions address different choices. One concerns the pace at which labs build more powerful systems; the other concerns restrictions on people using AI. Altman’s interview makes clear that supporting a slowdown does not, in his view, require accepting tightly concentrated control over the technology’s uses and benefits.