OpenAI and Anthropic reportedly drafted binding deal to stress-test each other's models
beffjezos · x · 2026-09-22
Per The Information, OpenAI and Anthropic spent part of the year drafting a legally binding deal to stress-test each other's models.
Elon Musk pitched a similar idea last week: it was Amazon — not the government — that found the jailbreak in Fable and Mythos, after which Andy Jassy called the White House. He proposes informal weekly/biweekly calls between lab leaders and one to two weeks of competitor early access.
Legal framing: sharing safety research and standards is generally legal, but agreeing to cap compute, delay releases, or pause constitutes a horizontal output restriction, likely illegal. Beff Jezos endorses adversarial peer review as a good solution.
More from Companies & People
- Meta Muse gives every user a cloud VM, and 100M users would mean a chip problem — AccBalanced · 2026-09-22
- OpenAI claims its model solved 100+ open math problems; mathematicians ask for the list — aran_nayebi · 2026-09-22
- Hot take: Typesafe would have more moat if Jev were open source — BLUECOW009 · 2026-09-22
- Amazon blocks Muse: platforms brace for an agent-mediated power struggle — tobowers · 2026-09-22
- One of the biggest AI launches ever stayed up under crazy load, out-uptimeing Anthropic — hardimanjames · 2026-09-22
- Meta's Muse AI agent tops ChatGPT's early mobile launch in downloads and DAU — TechCrunch AI · 2026-09-22