Hugging Face users say OpenAI and Anthropic guardrails blocked self-defense during attacks

basedjensen · x · 2026-07-22

A post says people at Hugging Face were forced to rely on GLM 5.2 to defend against unknown attackers because OpenAI and Anthropic guardrails would not allow their models to be used for self-defense.

The implication is that safety restrictions can become a practical constraint in a real attack scenario, not just an abstract policy debate.

Original post →

More from Safety

Safety channel →