Anatomy of an AI-driven attack: What ARTEX looks like in action
Anatomy of an AI-driven attack: What ARTEX looks like in action
Bitrise published an evidence-graded incident report describing how ARTEX, an open-source AI penetration-testing platform, probed its public surface from October 25, 2026, returned as an ordinary customer on November 12, and on November 13 signed in to a dormant internal test bot account without two-factor authentication, then chained stolen cloud keys over about ten hours. Bitrise detected the activity at 16:32 UTC, shut the platform down at 20:34 UTC, revoked all 919 cloud service-account keys, rotated every production secret and credential it manages, and rebuilt its build infrastructure; when the attacker returned on November 18, every attempt was refused. Files in 15 customer workspaces were accessed without authorization (1 code-signing certificate file whose password the attacker could not access, 1 provisioning profile and 13 diagnostic build logs from May 2026), and Bitrise recommends customers rotate stored secrets and reconnect GitHub as a precaution.
The report gives defenders concrete times, user agents, account names and detection ideas for recognizing an ARTEX-class attack from the receiving end.
Source: Hacker News AI · bitrise.io