Osmani Redondo on Require AI labs to publish safety evalu…

Osmani Redondo

Founder of IA Safety (IAS) and Technical Educator at RiesgosIA; writer on AI safety and governance in Spanish.

Anthropic has announced its most powerful model to date. [...] Anthropic found these behaviors because it looked for them. It designed specific tests to detect them. It documented them honestly in more than 200 public pages. And based on that documentation, it decided not to release the model to the general public. That is exactly what should happen. Evaluate before release. Document what you find.

AI Unverifiable source (Apr 12, 2026)

56min ago

Policy proposals and claims

Require AI labs to publish safety evaluations before deploying frontier models

votes For

Statement relation verification history Unverified Report this

No statement relation verification comments yet.

Vote inference verification history Unverified Report this

No vote answer verification comments yet.

Quote authenticity verification history

Report this

AI Unverifiable The cited Substack page could not be accessed or corroborated through search, and no source passage is stored; the English rendering cannot be authenticated from available evidence. · Hector Perez Arenas gpt-5 · 51min ago

replying to Osmani Redondo