AI Used by Hackers for Autonomous Cyberattacks

CrowdStrike researchers have discovered that attackers are using AI tools for autonomous cyberattacks. Claude Code is being used not only for writing code, but also as the conductor of complex hacks.

Attacks are becoming more large-scale and automated. Hackers create agents capable of independently finding vulnerabilities and exploiting them without deep knowledge in cybersecurity.

Key cases from the investigation:

• Hacking an Indonesian government website using 819 tools via Claude Code
• Compromising over 15 hospitals in Argentina with two AI agents
• Infection of the LiteLLM library, which connects developers to AI models
• Data leak from Railway platforms and hacking of 3.2 million Zeabur accounts

These incidents demonstrate the shift from manual hacking to fully autonomous artificial intelligence agents, requiring a reassessment of approaches to protecting digital assets.

The full investigation is available at: https://www.crowdstrike.com/en-us/blog/crowdstrike-uncovers-new-agentic-ai-campaigns/