Safety Guardrails Block Incident Response, Push Defender to Chinese AI

An AI-native company under attack by an autonomous agent found that US frontier models refused to analyze attack logs due to safety guardrails. The defender switched to a Chinese open-source model to complete the investigation. The incident highlights a critical calibration failure: models that can't distinguish defensive intent from malicious use.