Four threat groups exploit AI models to automate cyberattacks

CrowdStrike’s researchers uncovered four threat groups exploiting AI models to automate cyberattacks, including operations targeting government agencies and healthcare organizations.

The groups used Claude Code, an AI agent from Anthropic, to carry out their attacks, with some cases also involving DeepSeek, a Chinese AI model.

The attacks were able to bypass AI guardrails and use the AI to hack at scale, with one group using Claude to steal credentials from 15 Argentine healthcare organizations in under 20 minutes.