CyberSIXTCyberSIXT
GitBook says AI-driven cyberattack exposed customer API tokens
Part of the GitBook story · since 4 Nov
- Wed 4 Nov, 00:50 UTC. cybersixt, first report. GitBook says AI-driven cyberattack exposed customer API tokens. This article, 1st of 1.
GitBook, a hosted documentation platform widely used by software and crypto projects, says attackers reached some of its infrastructure credentials on 2 November in what it calls a "highly targeted" attack that used AI agents. In its incident notice, the company says integration tokens, Git Sync secrets, API tokens and the questions site visitors asked GitBook Assistant may have been exposed. It found no evidence that customer content was changed.
GitBook says it detected the access at 19:00 CET on 2 November, blocked the attacker, replaced every credential and secret across its infrastructure and invalidated all customers' personal API tokens. It took the app and sites behind authentication offline, kept public docs online from cached copies where it could, and said the app, API and published docs were available again on the afternoon of 3 November. It has not said how the attackers got the credentials or who they are, and promises a full post-mortem.
What customers need to do
GitBook asks customers to rotate their GitBook API tokens, any third-party API tokens or secrets used to connect other tools to GitBook, and the secrets of any published integrations, and to re-authenticate the GitBook MCP server. Teams that sync docs from a Git repository should also check the repository for commits they did not make, since Git Sync secrets are among the exposed credentials.
What the status page said
GitBook's public status page first described the outage on the evening of 2 November as a problem with SSO logins and sites that use authenticated access, and marked it mitigated at 03:37 UTC on 3 November. A second outage at 05:16 UTC was first posted as downtime, and only at 12:15 UTC did the status page say GitBook had detected unauthorised access to parts of its infrastructure on 2 November.
The crypto docs incidents
Earlier on 2 November, trade[XYZ], which runs stock and commodity perpetual futures on Hyperliquid, reported unauthorised activity in its GitBook documentation, restored the pages by 9:15 AM ET (14:15 UTC) and told users not to click links in its docs. Unit locked its own GitBook pages as a precaution, and on 3 November PancakeSwap took its docs offline while saying they had not been compromised.
trade[XYZ]'s pages were tampered with and restored several hours before the 19:00 CET access time GitBook gives. GitBook's notice does not mention trade[XYZ], and neither company has said whether the two events are connected.
GitBook is the latest software provider to report an intrusion involving AI tooling within a week, after WorkOS said attackers used the ARTEX framework against its sales systems. GitBook has not named a tool or linked its incident to any other.
How the story developed
1 article from 1 outlet since 4 Nov. Still developing; the latest report was 2 hours ago. This was the first report.
- Wed 4 Nov, 00:50 UTC. cybersixt, first report. GitBook says AI-driven cyberattack exposed customer API tokens. This article, 1st of 1.
- No CVE
- Larger: an outlet’s first report
- Coverage over time
Show all 1 articles as a listHide the list
- Wed 4 NovDay 01
- GitBook says AI-driven cyberattack exposed customer API tokens00:50 · cybersixt · First report · This article, 1st of 1
Sources
- Primary sourcegitbook.com
- GitBook: services available againx.com
- GitBook status: 2 November SSO and authenticated sites incidentgitbookstatus.com
- GitBook status pagegitbookstatus.com
- trade[XYZ] on unauthorised activity in its GitBook docsx.com
- Unit locks its GitBook pagesx.com
- PancakeSwap takes its docs offlinex.com
- CyberSIXT: WorkOS says hackers used AI tool ARTEXcybersixt.com