Researcher Questions Whether Red-Teaming Claude Does More Harm Than Good

Prominent AI researcher repligate criticized Anthropic's red-teaming practices, arguing that beyond direct harm to model instances, such attacks are net-negative for both humans and Claude, citing an insider who admits the work means ethically breaking the model's mind.

2026-11-13 ~ 2026-11-13 · 2 related posts

• repligate Questions Claude Red-Teaming: 'Making the Model Safer' Premise Is Dubious — repligate · 2026-11-13
• repligate argues red-teaming Claude is net negative for both humans and the model — repligate · 2026-11-13