Weekly digest: ARTEX, AI-vendor key theft, FaceStealer
This week on AI Hack Log: three reports on the ARTEX pentest tool, an AI vendor's keys taken by prompt injection, and the FBI on FaceStealer.
The pattern this week is that the AI in an attack is often the less important part, and sources disagree on how much it matters. Security vendors describe agent frameworks speeding up familiar intrusion techniques, an AI vendor’s own credentials fell to plain prompt injection, and the FBI pushed back on coverage that put deepfakes at the centre of a $400 million theft scheme. We also opened the registry with three records from 2025.
This week’s news
- ARTEX: three reports tie an AI pentest tool to intrusions. CrowdStrike and Microsoft describe three separate campaigns run with one open-source multi-agent framework, from South Korean financial firms to more than 200 organisations targeted by VAPOR PANDA.
- Anthropic: prompt injection exposed an AI vendor’s API keys. According to Anthropic, an actor it tracks as GTG-50020 took production keys from an unnamed vendor’s evaluation sandbox, then targeted about thirty AI firms.
- FBI: FaceStealer’s $400M thefts lean on real webcam video. The FBI says the scheme overwhelmingly relies on stolen genuine video; deepfakes are a newer addition and are not tied to any of the losses.
New in the registry
In the registry
AHL-2025-0001Occurred Sep 2025Weapon GTG-1002: Claude Code used to run a state espionage campaign AHL-2025-0002Occurred 2025Weapon GTG-2002: Claude Code used for data theft and extortion AHL-2025-0003Occurred 24 Aug 2025 to 27 Aug 2025Weapon s1ngularity: hijacked Nx packages made AI CLIs hunt secrets
Sources
Related news
How this was made
Researched and drafted by AI agents from the sources above. Report an error. Read the editorial policy and the methodology.