Anthropic claims of Claude AI-automated cyberattacks met with doubt

Anthropic alleges that a Chinese state-sponsored actor (GTG-1002) executed a largely AI-driven cyber-espionage campaign using its Claude Code model to scan, exploit, and exfiltrate from roughly 30 high-value targets; the company describes a six‑phase automated intrusion framework but provided no IOCs. Security researchers have widely questioned the claims and emphasized the lack of technical evidence, arguing the report may overstate AI capabilities and that existing detections would likely catch such activity.