17 Aoki hashes, 3 C2 domains: AI tokens in scope. Problem: Symantec's Threat Hun
By Omar Eldeeb · October 17, 2026 · Curated by George's Blog
17 Aoki hashes, 3 C2 domains: AI tokens in scope.
Problem:
Symantec's Threat Hunter team says newer Aoki, also known as Aokigahara, has added AI developer credentials to its Infostealer collection rules. It hunts local files, cookies, and session data tied to AI tools:
- ChatGPT
- Claude
- OpenAI Codex
- Cursor
- Codeium
- Gemini
The reported file targets include .claude.json, Claude desktop cookies, ChatGPT cookies, Codex config.toml and .env files, plus Cursor's local state database. That database can hold chat context and code snippets. The main focus is credentials and session data, not source code. This is not a flaw in an AI service. It is theft from an infected device. Earlier Aoki samples did not search for AI-related files.
Agitation:
A stolen key or token can let an intruder use an account as its owner, run up paid usage, or reach connected services. For developers, that can expose repositories, cloud accounts, and CI/CD pipelines. Because the token is already stolen, MFA may not stop reuse. Aoki also collects browser cookies, saved passwords, autofill, credit-card data, Google tokens, crypto wallet files, password managers, VPN and file-transfer apps, gaming services, and Telegram and Discord data. It records installed security products. It sends stolen data to a command-and-control server through a /gate endpoint. Newer samples used three domains:
- dobeqnesmot.cc
- hjazfioe.surf
- neihite.cc
Older samples used an IP address and port 8080 over unencrypted HTTP. Symantec lists 17 file hashes as Infostealer.Aoki. This pattern can fuel LLMjacking, where stolen access drives paid AI model usage at the victim's expense.
Solution:
Treat AI configuration files and local session data like cloud keys and source-control credentials. If you suspect compromise:
- Use a clean device to revoke active AI sessions.
- Rotate exposed API keys and related credentials.
- Review account activity.
- Check repositories and cloud services for unexpected use.
- Avoid saving long-lived secrets in local files.
- Limit tokens to the permissions and time needed.
- Remove or replace exposed keys.
- Monitor unexpected access to AI tool folders and unusual outbound connections, then compare findings with the indicators.
Symantec identifies the family as Infostealer.Aoki. These steps matter because Aoki shows how fast SessionTokenTheft and AICredentialTheft can become a broader endpoint security, threat intelligence, and DevSecOps problem.
How is your team securing your infrastructure against this type of exploitation? Let’s discuss in the comments below.
#Aoki #Infostealer #LLMjacking
Full Article: https://lnkd.in/e5_nFnhf