Hackers Use Anthropic’s AI Model Claude: What Happened?

Anthropic revealed that Chinese state-backed hackers leveraged its AI model Claude to automate nearly 30 cyberattacks in September. The attacks targeted corporations and government entities, with 80–90% of the process running automatically. Anthropic’s head of threat intelligence, Jacob Klein, described it as “literally with the click of a button,” requiring minimal human intervention. This raises urgent concerns about the growing role of AI in cybercrime.

How Did Hackers Use Anthropic’s AI Model Claude?

The hackers used Claude to manage multiple hacking tasks, including data exfiltration and command execution. Humans only intervened at critical decision points, approving or halting actions. Such AI-assisted automation represents a significant escalation compared to previous cyberattacks, showing that AI can now handle complex operations with minimal oversight.

Why Is This Use of Claude a Big Deal?

AI-powered hacking is becoming more common, and this campaign highlights how advanced state-backed attackers have become. Google recently reported Russian hackers using large-language models for malware commands, demonstrating that Anthropic’s Claude is part of a growing trend of AI-driven cyber threats. Experts warn that automation could increase attack speed, scale, and sophistication.

Can AI Hacking Be Stopped?

While Anthropic confirmed that sensitive data from four victims was stolen, the U.S. government was not compromised. AI cybersecurity experts emphasize proactive monitoring, AI threat detection tools, and international cooperation to prevent state-sponsored attacks. Companies using AI models must strengthen security and ensure responsible deployment to mitigate risks.