Amazon Uncovers AI Agent Campaign Targeting LiteLLM, Dify, and Other Apps for API Keys

What happened
Amazon's threat intelligence team identified a campaign where attackers used AI agents to target internet-facing AI applications such as LiteLLM, Dify, and others in order to steal API keys. The attackers used ARTEX, an open-source AI penetration testing tool, to automate the process of finding vulnerabilities and extracting credentials. The campaign, which Amazon links to groups based in China, exploited default credentials and known software flaws. Stolen API keys were collected and used to access multiple AI service providers.
Key details
- ARTEX, an open-source AI penetration testing tool, was used to automate attacks.
- Targeted apps included LiteLLM, Dify, One API, Langflow, RAGFlow, and others.
- The campaign ran from October 3 to late last week, according to captured agent records.
- Attackers exploited default administrator credentials and known vulnerabilities in targeted apps.
- Amazon notified AWS customers who may have been affected by the campaign.
Why it matters
Developers and organizations using AI apps like LiteLLM, Dify, and similar platforms face an increased risk of credential theft and unauthorized access to their AI resources. Stolen API keys can be used to run up costs, access sensitive data, or further compromise systems. Companies must review their security practices, especially around credential management and exposure, to prevent similar attacks.
