Ledger / Oct 2026
ARTEX agents run a self-funding campaign stealing AI-model credentials, seen through AWS MadPot decoys
Confirmed
Amazon read the agents' own records: 'the agent records show that agentic frameworks have progressed ... to sustained, multi-objective offensive operations'. The framework's identity is hedged ('assesses with moderate confidence that ARTEX was used'); the agent use is not.
- Model
- Not namedARTEX, real-time face swap (model not named)
- Agent harness
- ARTEXopen-source harness
- Attributed country
- Not attributedthe source names no country
- Sectors hit
- ai platforms, cryptoOwners of exposed AI-service credentials; operators' stolen keys paid for further…
Operators ran ARTEX agents on standing instructions to hunt AI-model credentials, using stolen keys to pay for the inference, and to attempt KYC bypass at cryptocurrency exchanges with real-time deepfake injection. Amazon observed this from 3 October to 5 November 2026.
“The most prominent ARTEX campaign was a large-scale theft of credentials to access AI models, organized as a self-sustaining operation in which stolen keys paid for the LLM inference powering further credential hunting.”Amazon
Sources
- Date
- Oct 2026
- Actor
- Unnamed cybercriminal operators using ARTEX (criminal), as attributed by Amazon
- Targets
- Owners of exposed AI-service credentials; operators' stolen keys paid for further inference
- What the AI did
- Agentic intrusion, LLMjacking, Deepfake, Fraud
- Model as named
- real-time face swap (model not named)
- Model basis
- A researcher names it
- Autonomy
- Semi-autonomous
- Outcome
- Stolen AI-model credentials funded the inference that ran the campaign; crypto exchanges were targeted for KYC bypass.