AI Agents Enable Criminal Cyberattack Automation

The International AI Safety report, chaired by Yoshua Bengio and authored by over 100 experts across 30 countries, finds that AI systems increasingly assist in automating stages of cyberattacks. It cites Anthropic's November 2025 disclosure of Chinese actors using Claude Code against about 30 organizations, DARPA's AIxCC finding that finalist models identified 77% of synthetic vulnerabilities, and notes criminals trade weaponized malware models for roughly $50 per month.