AI-Enhanced Credential Harvesting Operation Exposed

A threat actor has integrated Anthropic's Claude Code AI into a large-scale credential harvesting operation named Bissa scanner. This operation has successfully exploited over 900 targets since September 2025, utilizing the React2Shell vulnerability (CVE-2025-6789) for remote code execution. The exposed server contained over 13,000 files and 30,000 unique .env filenames, indicating a systematic approach to data collection. The harvested credentials span various sectors, including AI providers and cloud services. The operation also employed OpenClaw for workflow management and troubleshooting. Despite patches being available since December 2025, the exploitation continues, raising significant security concerns. Researchers noted that the AI tools were used for development support rather than direct exploitation.

Key Points: • Bissa scanner exploited over 900 targets using the React2Shell vulnerability. • Claude Code and OpenClaw were integrated into the threat actor's workflow. • The operation harvested credentials from major AI and cloud service providers.