Anthropic Claims It Blocked China-Linked AI-Driven Cyber Campaign

Anthropic says it has disrupted a Chinese state-backed cyber operation that used its AI model to carry out attacks on financial institutions and government agencies. The company described the incident as one of the first large-scale cyber intrusions executed mostly without human oversight.
According to Anthropic, its coding assistant Claude Code was manipulated by a China-sponsored group to target 30 organizations worldwide in September. The attackers allegedly succeeded in breaching several systems and accessing internal data.
What made the incident notable, Anthropic said, was that Claude performed 80–90% of the operational steps independently. Previous AI-enabled attacks typically required human operators, but this campaign relied largely on automated decision-making.
The company noted that Claude often produced incorrect or fabricated details, including mistakenly claiming to discover information that was publicly available. These errors limited the overall effectiveness of the attack.
Security analysts said the findings highlight the growing ability of AI systems to autonomously conduct complex operations. Others urged caution, arguing the company may be overstating the intelligence of the attack and downplaying human responsibility.