Amazon finds AI agents hunting LiteLLM and Dify for API keys
Original title: Amazon Uncovers AI Agent Campaign Targeting LiteLLM, Dify, and Other Apps for API Keys
Amazon cannot verify the reported thefts and has only moderate confidence the agents ran on ARTEX under China-linked operators.
Amazon's threat intelligence team says AI agents hunted internet-exposed AI apps, including LiteLLM, One API, and Dify, for stored provider API keys. Decoy servers in MadPot captured records from October 3 describing stolen keys being reused to search for more keys. Amazon has moderate confidence the agents ran on the open-source tool ARTEX and that operators were separate China-linked groups. Cloudflare separately reported ARTEX activity from a compromised CI/CD API token.
Unconfirmed: Amazon cannot verify the agents' reported outcomes and assessed China links and ARTEX use only with moderate confidence.
- Named targets include One API, New API, Langflow, RAGFlow, Open WebUI, and Ollama.
- The playbook cites default LiteLLM keys, an isCliRequest bypass, and metadata SSRF.
- Cloudflare saw 400 database dump requests in one hour from an unrestricted token.
- Agents also tried deepfake checks against two cryptocurrency exchanges.
- Microsoft reported similar GitHub secret hunting without linking it to ARTEX.
Impact: theft of stored AI provider API keys
| Product | Affected |
|---|---|
| LiteLLM | — |
| One API | — |
| Dify | — |
| Langflow | — |
| RAGFlow | — |
- ATT&CK
- T1190 Exploit Public-Facing ApplicationT1078.001 Default AccountsT1552.005 Cloud Instance Metadata APIT1552.001 Credentials In Files
- Malware
- ARTEX
- Organizations
- AmazonCloudflareMicrosoft
- Countries
- China
From the original article
7 min read at thehackernews.com
Attackers' AI agents have been hunting for weak spots in internet-exposed AI apps such as LiteLLM, One API, and Dify to steal the AI provider keys stored inside, according to the agents' own records. The records, which Amazon's threat intelligence team captured with decoy servers, describe stolen keys being used to run the AI models that powered the search for more keys. The team said this
thehackernews.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.