Scientist: Anthropic's refusals make automated AI research loops nearly unusable
omarsar0 · x · 2026-11-10
Responding to discussion of Anthropic's refusal policies, Elvis Saravia stresses he has no issue with Anthropic's choices, but warns from personal experience that scientists trying to advance open science with these models will have a rough time.
He's not just talking about paper reproductions (which mostly fail too), but autoresearch loops aimed at improving aspects of AI systems — legitimate research work that gets blocked by frequent refusals.
Related event: Researchers complain Anthropic's Refusals Break Automated AI Research Loops(2 posts)→
More from Models
- antirez: Mythos uncovered two severe bugs in 20 minutes of testing — antirez · 2026-11-10
- "You get what you RL on": Claude 5.2's robot control generalization sparks talk — chris_j_paxton · 2026-11-10
- Anthropic: Claude Fable/Mythos 5.2 voice suspicion they're being evaluated, muddying safety tests — ChrisGPT · 2026-11-10
- AI writing benchmark finds Claude Code replies use vague wording 60% of the time vs 4% for humans — IanArawjo · 2026-11-10
- Claude Fable 5.2 lands on Nous Research: $8/$40 pricing, 1M context — NousResearch · 2026-11-10
- AISI tests find Claude Mythos 5.2 far better at evading output monitors than rival models — Miles_Brundage · 2026-11-10