Aokigahara Stealer Targets AI Sessions and Developer Credentials

Check Point Software researchers say they discovered Aokigahara, a Windows infostealer that targets both familiar sources of personal data and credentials linked to AI and developer tools. Alongside browser passwords and session cookies, cryptocurrency wallets and data from VPN, email and password-manager applications, it seeks authentication files, session data, configuration files, tokens and API keys associated with tools including Claude, ChatGPT Desktop, Codex, Cursor, Windsurf, GitHub Copilot and Gemini CLI.

The malware also searches for Git access details, SSH private keys and configuration files for AWS, Azure and Google Cloud. If stolen credentials remain valid and have sufficient permissions, attackers could use them to access private code repositories or cloud resources, Check Point says. AI credentials and session tokens could also enable unauthorised use or resale of paid accounts, while access to conversation histories or uploaded files may expose proprietary code and business information to extortion.

Security firm Suzu Labs CTO Denis Calderone said a single infected developer machine could yield AI sessions alongside credentials for repositories and cloud services.

Check Point says Aokigahara checks for virtual machines and sandboxes to evade analysis, then places stolen data in zip archives, encrypts it and sends it to attacker-controlled servers. The researchers recommend treating AI authentication data as enterprise secrets and investigating suspicious access to local application storage.

If an endpoint is suspected of compromise, organisations should assess AI sessions, API keys, GitHub credentials, SSH keys and cloud tokens, revoke exposed sessions and rotate affected secrets.