{"slug": "amazon-s-decoys-capture-ai-agents-claims-of-stolen-key-powered-hunts", "title": "Amazon's decoys capture AI agents' claims of stolen key-powered hunts", "summary": "Amazon Threat Intelligence reported on November 13 that its MadPot decoy sensors, configured to resemble model endpoints, captured agent records from October 3 to November 5 in which AI agents hunting for exposed inference gateways sent requests carrying their working context — task histories, tool records, system prompts and notes on prior findings. Amazon said it could not verify the outcomes those records reported.", "body_md": "# Amazon's decoys capture AI agents' claims of stolen key-powered hunts\n\n**Amazon's November 13th account says MadPot captured agent records from October 3rd to November 5th describing hunts through exposed AI gateways. Amazon could not verify the outcomes those records reported.**\n\n        By [RuntimeWire Staff](https://runtimewire.com/author/runtimewire-staff)\n        · Published \n\nAmazon's MadPot decoys were set up to look like model endpoints. Agents looking for inference mistook the sensors for the real thing, Amazon Threat Intelligence said in its November 13th account, and sent requests that carried their working context with them - task histories, tool records, system prompts and their own notes on what they had found elsewhere.…", "url": "https://wpnews.pro/news/amazon-s-decoys-capture-ai-agents-claims-of-stolen-key-powered-hunts", "canonical_source": "https://runtimewire.com/article/attackers-are-using-stolen-ai-api-keys-to-pay-for-agents-that-hunt-for-more-keys", "published_at": "2026-11-17 22:34:11+00:00", "updated_at": "2026-11-17 22:42:40.014548+00:00", "lang": "en", "topics": ["ai-safety", "ai-agents", "ai-infrastructure", "ai-policy"], "entities": ["Amazon", "Amazon Threat Intelligence", "MadPot"], "also_reported_by": [], "takeaway": "Agents probing for exposed model endpoints leaked their own system prompts, tool records and task histories to Amazon's MadPot decoys between October 3 and November 5, so treat any unverified inference gateway as a sensor that captures agent context.", "recommendation": {"advice": "Treat unverified or exposed model endpoints as hostile sensors: scope agent credentials, strip secrets from system prompts and tool records, and restrict egress so probing agents cannot leak their working context.", "playbook": "agent-security", "playbook_url": "https://wpnews.pro/playbooks/agent-security"}, "alternates": {"html": "https://wpnews.pro/news/amazon-s-decoys-capture-ai-agents-claims-of-stolen-key-powered-hunts", "markdown": "https://wpnews.pro/news/amazon-s-decoys-capture-ai-agents-claims-of-stolen-key-powered-hunts.md", "text": "https://wpnews.pro/news/amazon-s-decoys-capture-ai-agents-claims-of-stolen-key-powered-hunts.txt", "jsonld": "https://wpnews.pro/news/amazon-s-decoys-capture-ai-agents-claims-of-stolen-key-powered-hunts.jsonld"}}