GitBook Hit by AI‑Agent Breach, Urges Customers to Rotate Keys
Researchers say the attack highlights how automated AI agents can be used to harvest credentials and threaten connected services.
Overview
- GitBook says a targeted intrusion using AI agents gained unauthorized access to some infrastructure credentials on Nov. 2 and the company detected and blocked the attacker quickly.
- As part of containment, GitBook took the app and authenticated docs offline, replaced every credential and secret across its infrastructure, and revoked all personal API tokens.
- The company’s investigation found the attacker may have accessed Git Sync secrets, GitBook API tokens, third‑party integration tokens, and questions sent to the GitBook Assistant, with no evidence so far that hosted content was changed.
- GitBook has restored its app, API and published docs while warning some services may remain degraded and instructing customers to rotate GitBook tokens, rotate third‑party secrets, and reauthenticate MCP server connections.
- Security researchers at Google TIG and Mandiant have recently documented similar AI‑agent tooling in real intrusions, which raises the risk that compromised credentials could be reused to move between systems or extract data from linked services.