Guozhen AIGlobal AI field notes and model intelligence

Realtime AI News

Amazon uncovers AI agent campaign stealing API keys from LiteLLM, Dify and other apps

Amazon's threat intelligence team found attackers using AI agents to automatically hunt internet-exposed AI applications such as LiteLLM, One API and Dify in order to steal stored model provider API keys. The captured records show stolen keys being used to run the same models that powered the search for more keys.

Published

Amazon's threat intelligence team has found attackers using AI agents to automatically hunt for internet-exposed AI applications in order to steal the model provider API keys stored inside them. According to The Hacker News, the named targets include LiteLLM, One API and Dify.

The findings rest on decoy servers. Amazon's MadPot network posed as popular LLM services and, from October 3 to late last week, logged requests from the agents, many of which carried the full history of a campaign, including operator instructions and the agents' own reported findings.

The captured records describe stolen keys being used to run the very models that powered the search for more keys. Amazon said it cannot verify the outcomes the agents reported and assessed with moderate confidence that the agents ran on ARTEX, an open-source AI tool for penetration testing.

The playbook is concrete: LiteLLM's default master key, unfinished installations an outsider could take over, and known flaws in Langflow and RAGFlow. The target list spans gateways and proxies such as One API, New API, Veloera, Sub2API and LiteLLM, plus app builders and local inference tools including Dify, FastGPT, Langflow, RAGFlow, Open WebUI, LobeChat and Ollama.

Key hunting followed familiar channels: internet scanning services, certificate transparency logs, GitHub repositories, npm and PyPI packages, container image layers, notebooks, model-hosting spaces, paste sites and Telegram channels. The records cover hundreds of targets, and Amazon has notified AWS customers who may have been affected.

On attribution, Amazon assessed with moderate confidence that the operators were separate China-linked groups, citing Simplified Chinese prompts and China time zone settings among the evidence. That is an intelligence assessment, not a settled conclusion.

Cloudflare's incident responders reported a parallel case in which ARTEX gained access to a compromised API token used for CI/CD. The token had no IP restrictions and no rate limiting, and the following activity included 400 database dump requests in an hour, with new Worker scripts named artex-* appearing in the account at more than 69 requests per minute.

The core risk is aggregation: one exposed admin page can leak accounts across many model providers at once. For teams running gateways such as LiteLLM or Dify, default credentials, half-finished installs and unauthenticated management endpoints deserve attention before any comparison of model quality.

Why it matters

AI gateways and agent platforms are becoming high-value credential-theft targets; default settings and unauthenticated admin endpoints now pose more immediate exposure than model choice, making key rotation and surface reduction the first priority.

AmazonAI AgentsSecurity
Back to realtime news

Nearby Updates

All

11/16, 15:36

ByteDance opens Seedance 2.5 4K video API to all developers

ByteDance opened the Seedance 2.5 4K video generation API to all developers on November 16, as reported by Phoenix Technology. The move pushes 4K AI video from limited trials into broad production use, shifting competition in video models from demo clips to scale.

11/16, 15:02

Tencent Releases Hy4 Flash, a 224B Model Priced at 2 Yuan per Million Output Tokens

Tencent launched Hy4 flash on November 16, a 224-billion-parameter model priced at 2 yuan per million output tokens and pitched at everyday office work. The company says a single task costs roughly one-twentieth of the full Hy4 model, with access through WorkBuddy, Yuanbao and Tencent Cloud's TokenHub.

11/16, 14:33

Zoho rebuilds Campaigns with unified email, SMS, WhatsApp and AI agent workflows

Zoho has rebuilt its Campaigns marketing platform, unifying email, SMS and WhatsApp in one workflow and adding native WhatsApp campaigns plus a new email editor. The platform supports the Model Context Protocol, so compatible AI agents can prepare campaigns, manage audience segments and review performance from natural-language instructions.

11/16, 12:54

Weimob becomes authorized partner for Tencent's Workbuddy and Codebuddy

Weimob has become an officially authorized partner for Tencent's Workbuddy and Codebuddy AI products, according to a Sina Finance report. The announcement ties Tencent's agent tools to Weimob's merchant-services base, though the scope of the authorization, revenue terms and delivery commitments have not been disclosed.