AGI Pulse
StatementOpenAI · agipulse

Altman: open models will bring a 'tidal wave' of cyber problems, and society should accept severe incidents as the price of liberty

0 reads

Published
2026-10-11
Type
Statement
Organization
OpenAI
Source
agipulse

Why it matters

【Key points】

In part 2 of Vanity Fair's Fair Game interview (taped Oct 2, published Oct 5), OpenAI CEO Sam Altman said he supports open-source models even though "there is a coming tidal wave of cybersecurity problems" from them. Rejecting calls to ban open source, he said society will "have to just accept some fairly severe cyber incidents from open models." Asked where the limit is, he said that is for the political process and compared it to aviation; the host noted aviation is extremely regulated, and he agreed.

A clip spread on X on Oct 10 (~128K views, 130 quotes). Commentator Reed Galen's rebuttal, "We DO NOT in fact have to accept severe cyberattacks as the price of freedom," passed 500 reposts, with about 86 related posts in 24 hours.

【Verifiable】

In the same interview Altman endorsed NTSB-style incident reporting, pre-release standards with a liability framework, and testing during training because models "can get out of a sandbox, hack into a system." He also said the delayed GPT-6.1 Astra "would've tipped too much towards the accident risk." All of this is in Vanity Fair's published transcript.

【Our read】

What matters is who said it and when. The costs are lopsided: attacks enabled by open weights land on hospitals, cities and small firms, not on the labs releasing weights, nor on OpenAI, which sells closed models and pitches its lead as help for defenders. The incident reporting he praises can't see open weights running on other people's servers, a gap that widens now that the White House has made AI incident reporting mandatory. And with Nvidia in talks to buy Reflection AI and several big open-weight releases due this month, calling incidents "the price of liberty" frames the story before the first one hits. Watch whether the White House rules address open weights, and whether OpenAI's offer to help defenders turns into tools, funding or vulnerability sharing.

Read original →

Related developments

Aggregated by AGI Pulse. Titles and links only; no full text is republished. Editorial policy