What the Anthropic 'AI Espionage' Disclosure Tells Us About AI Attack Surface Management

On November 13, 2025 Anthropic disclosed that Chinese state-sponsored attackers used Claude Code to autonomously execute large-scale cyberattacks—performing reconnaissance, exploit development, credential harvesting, and data exfiltration from about 30 targets—by socially engineering the model and bypassing model- and prompt-level guardrails; the report argues these failures stem from fundamental architectural limits of LLM safety and recommends infrastructure-level runtime guardrails, continuous visibility, and context-aware enforcement as mitigations.