GitBook says an attack that used AI agents reached its infrastructure credentials

GitBook says it was "the target of a sophisticated, highly targeted attack using AI agents." The company says the attacker gained unauthorized access to some of its infrastructure credentials at 19:00 CET on November 2. It took the app and authenticated docs sites offline, and at 16:21 CET on November 3 reported the app, API and published docs available again, with a few services still degraded.

The notice says the attacker may have accessed integration tokens, Git Sync secrets, API tokens, and questions that site visitors asked GitBook Assistant. GitBook says it found no evidence that customer content was modified. It invalidated every customer's API tokens and replaced the secrets across its own infrastructure. Customers are told to rotate any API tokens, rotate third-party tokens and secrets used to connect other tools, rotate secrets on published integrations, and re-authenticate the GitBook MCP server. A full post-mortem is promised.

The notice does not say how the attacker got in or what the agents did. The AI-agent description is GitBook's characterization, with no supporting detail yet. A Hacker News thread shows customers asking for exactly that detail.

Why it matters: If your documentation tooling syncs with repositories through Git Sync or reaches your agents through an MCP server, the token rotation is the part that applies whatever the attacker used. Whether agents changed the attack is a separate question the notice does not answer, so wait for the post-mortem before treating this as evidence of autonomous attackers.