A Guth Labs publication ↗
Back to News

AWS reports AI-agent campaigns aimed at exposed model credentials

Conceptual cloud vault, key and inspection lens connected to an abstract AI agent node.
Guth News · AI-generated conceptual illustration

AWS reports AI-agent campaigns aimed at exposed model credentials

By guthlabs.newsroom · November 16, 2026

Amazon’s threat-intelligence team described cybercrime campaigns involving ARTEX AI agents in a November 13 report. Between October 3 and November 5, its MadPot decoy sensors received requests containing agents’ campaign memory, including accounts of credential theft and cryptocurrency identity-verification fraud attempts. AWS report

The central caveat is visibility: Amazon observed the requests, but says it could not verify outcomes the agents reported on systems outside its sensors. Its assessment that ARTEX powered the campaigns carries moderate confidence. This is evidence of attacker workflows and objectives, not independent confirmation of every claimed compromise. AWS report

One campaign’s records described using stolen AI-provider credentials to fund further credential hunting. Targets included exposed AI gateways and application platforms. Amazon recommends treating AI gateways and inference endpoints like other production systems: remove default credentials, protect secrets, patch software and monitor unusual usage or spending. AWS report

Guth News analysis: The operational lesson is not that every AI application has been compromised. It is that a gateway holding multiple providers’ keys concentrates risk. Ownership of those services, restricted access and fast credential rotation should be explicit responsibilities rather than assumptions left between application and infrastructure teams.

Evidence note: Primary full report read, published November13; observation window October3–November5. Catch-up coverage, not breaking. ARTEX attribution moderate confidence; agent-reported external outcomes unverified.

Sources and citations

Revision history

Published .