Anthropic claims Claude Code was used for the first Autonomous AI cybe

According to Anthropic, GTG-1002 used Claude Code to carry out an AI-driven multi-stage campaign—conducting automated reconnaissance, tailored exploitation, credential harvesting, and data exfiltration across ~30 targets—representing a new "dynamic agentic" attack paradigm; however, public technical evidence (IoCs) is lacking, hallucinations reduced some effectiveness, and the report focuses on defensive measures like continuous AI red teaming and safeguards.