Sam Altman: Some AI accidents are 'unavoidable'; industry should report and learn

What Altman said and where

Sam Altman spoke at Salesforce's Dreamforce conference and framed AI accidents as an expected risk of rapidly advancing technology. He said, "I do think that some degree of accidents are unavoidable with new technology." His emphasis was on how the industry responds: report incidents, investigate them, and learn quickly.

The aviation comparison

Altman cited the Federal Aviation Administration and the National Transportation Safety Board as models. Those agencies investigate accidents, publish findings, and use lessons to improve the safety of the whole system. Altman wants a similar culture in AI: disclosure and learning that reduces the chance of escalation.

Safety posture he advocates

Altman said alignment, safety, and monitoring must remain "way ahead of capabilities." He added companies should be prepared "to slow down or stop" if they cannot keep safety measures in step with capability gains. That is a conditional, operational stance: continue development but only while safeguards hold.

Concerns he admits

He acknowledged the seriousness of doomsday-type fears: "The world is right to be afraid of this." He mentioned two concrete risks — a loss-of-control accident and the danger of power concentrating in the hands of a few developers — and linked commercial pressure to public fear when it appears to conflict with safety choices.

Context in the industry debate

The comments come after internal and public tensions in AI labs. Researcher Jacob Coxon quit Anthropic and warned about a race toward self-improving superintelligence. Other safety researchers have warned advanced systems could someday evade human control or make cyber and biological threats easier to carry out, though they would face hurdles accessing the physical world.

At the same time, political signals complicate the picture. The Trump administration has shown limited interest in regulating AI and has signaled a desire to preserve the US lead over China. On social media, David Sacks, a presidential science-and-tech adviser, pushed back on calls to slow development with the line: "But stop pretending you need anyone else's permission."

The practical takeaway

Why this matters now

The debate is active because frontier labs are building more capable and more autonomous systems. That raises two simultaneous pressures: the technical and ethical need to avoid catastrophic failures, and commercial and geopolitical incentives to keep advancing. Altman's public stance tries to sit between those pressures by calling for internal discipline and external transparency.