Vercel breach traced to compromised Context.ai OAuth app with access to an employee's Google Workspace

Vercel, the cloud platform behind Next.js, disclosed on April 19 that attackers had reached some of its internal systems and a limited set of customer credentials. The entry point was a third-party AI tool, not a flaw in Vercel’s own products. Vercel warned that the same compromise may reach hundreds of users across many organisations.

What happened

A Vercel employee signed up for Context.ai’s consumer “AI Office Suite” with a corporate Google Workspace account and approved broad OAuth permissions. Context.ai later confirmed that an attacker had breached it in March and probably stole OAuth tokens belonging to consumer users. The attacker used the Vercel employee’s token to take over that Google account. From there they moved into Vercel’s internal systems, where some credentials were stored unencrypted.

Researchers who analysed the case say the Context.ai compromise itself started with an infostealer infection on a Context.ai employee’s machine. That would make this a chain from one stolen-credential event to several downstream breaches.

Impact

  • The exposed data was customer environment variables that had not been marked “sensitive” in Vercel. Vercel contacted the affected customers directly.
  • A forum seller offered what they described as Vercel customer API keys, source code and database data. The listing claimed a link to ShinyHunters, but the group denied involvement to BleepingComputer.
  • Vercel said its open-source Next.js and Turbopack projects were not affected.
  • The compromised Context.ai OAuth client was published so that other Google Workspace administrators can search for it.

Recommendations

  • Rotate every secret stored in Vercel environment variables that was not marked sensitive, and mark it sensitive from now on.
  • In Google Workspace, review third-party app access and search for the Context.ai OAuth client ID (110671459871-30f1spbu0hptbs60cb4vsmv79i7bbvqj.apps.googleusercontent.com).
  • Limit which OAuth scopes users can grant to unverified or consumer apps. Require admin approval for apps that request full access.
  • Check source repositories and deployment logs for unexpected activity since March.
  • Tokens and session cookies taken by infostealers are a common root cause in cases like this. Checking corporate domains against leaked-credential sources such as TAYNOPIS can surface exposed accounts early.

Sources