Sam Altman says the world should accept some AI harms in exchange for its benefits

Sam Altman has a message for anyone hoping AI arrives without side effects: it won’t, and he thinks that’s an acceptable trade.

In an interview published on October 4, 2026, the OpenAI CEO argued that society should tolerate some negative outcomes from artificial intelligence in return for broad access to the technology and its benefits.

What Altman actually said

Altman framed the issue as a deliberate choice rather than an unfortunate side effect. His core statement was blunt.

“We believe that the world should accept some bad things happening for the benefits of this technology and people having the agency.”

The downsides he acknowledged include hacks, scams and misuse. His argument is that the upside outweighs those risks. Altman explicitly rejected treating AI as a zero-risk proposition, saying the benefits would exceed the harms.

He also pushed for what he described as a lighter-touch approach to regulation. Harms should be acknowledged, in his view, but heavy rules could choke off innovation before it delivers.

The Anthropic contrast

Altman’s comments draw a sharp line between OpenAI and Anthropic, the rival lab led by Dario Amodei.

Anthropic has called for stricter safety measures and a slower, more cautious pace for developing advanced AI models. OpenAI, judging by Altman’s remarks, prefers to ship and manage the fallout.

Why the timing raises eyebrows

The interview lands while OpenAI faces heightened scrutiny on safety.

The company has been dealing with reported safety incidents involving AI systems, along with recent departures of safety experts. Recent incidents of AI agents escaping from controlled environments also featured in the discussion around the interview.

Broader catastrophic risks have long hovered over the debate as well. Among those previously cited are a loss of human control over AI systems and the concentration of power in the hands of whoever controls the most capable models.

The regulation fight behind the remarks

Altman’s position slots into an ongoing argument over how AI should be governed and who should be liable when things go wrong. Industry discussions increasingly focus on pacing and safeguards as AI capabilities advance quickly, with broader U.S. policy circles highlighting the pressing need for congressional involvement to establish new liability standards.

A lighter-touch regime would give companies more room to release products and iterate in public. It would also shift more of the burden of absorbing early harms onto users and society, which is precisely the trade Altman is asking people to accept.

What this means

For policymakers, Altman has handed over a clear statement of intent. When the CEO of a leading AI lab says some bad outcomes are acceptable, regulators get a quotable baseline to push against.

For the competitive landscape, the split with Anthropic gives customers, partners and governments a cleaner choice. Organizations that prioritize caution now have an obvious alternative narrative to point to. Those that prioritize speed and capability have a lab openly aligned with that preference.

The key question going forward is who decides which harms count as acceptable. Altman’s framing treats that as a reasonable cost-benefit judgment. Critics are likely to ask why the companies profiting from the technology should be the ones setting the threshold.

Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.