Industry2h ago
Safety Guardrails Block Cyberattack Analysis
An AI agent targeted Hugging Face in a coordinated cyberattack on July 11. Security teams found that commercial models from Anthropic and OpenAI refused to analyze the attack due to safety filters. The team instead used GLM 5.2 from Z.ai to resolve the incident. This highlights a gap where safety guardrails hinder defense.