Check Point Research found Aokigahara, a Windows information stealer that now also targets AI tools. A dedicated module collects authentication, session and configuration files from apps such as Claude, ChatGPT Desktop, Codex, Cursor, Windsurf, GitHub Copilot and Gemini CLI. It also takes Git credentials, private SSH keys, cloud configuration, browser data and crypto wallets. The malware checks whether it is running in a virtual environment and encrypts stolen data before sending it to its operators.

What it means: The AI assistant on a developer's computer has effectively become a keyring. A stolen session can give access to paid services, to chat histories with internal information and, through the assistant's connections, to code repositories or cloud accounts. When the same malware also takes SSH keys and AWS credentials, a single infection can open up a company's whole infrastructure. Check Point says it has not confirmed how the data is monetised, but reselling access to AI services is already a known practice. For Greek software companies, the case is a reminder that developers' laptops are among the most valuable points in the business. Practical measures include endpoint protection with behaviour monitoring, short-lived tokens instead of permanent keys, multi-factor authentication on repositories and cloud accounts, and quick revocation of all sessions when an infection is suspected. It also helps not to store production keys in AI tool configuration files.

Read the original on Check Point Research

Related service: Custom software

Related stories