Vercel Breach: OAuth Attack Reveals Hidden Risk of Platform Environment Variables

Lumma Stealer infection on a Context.ai employee leaked Google Workspace OAuth tokens, which were exploited to access a Vercel employee account and internal systems, resulting in enumeration of environment variables in some customer projects.

The core risk was Vercel's method of storing non-sensitive environment variables. Attackers were able to access even variables without a sensitive marking, and if these contained downstream service credentials, it could lead to further breaches.

Key details are as follows.

• Infection began on the Context.ai side around February 2026
• Moved to a Vercel employee's Google Workspace account in March 2026
• Vercel's security notice and public explanation on April 19, 2026
• In at least one case, a public report stated that a customer received an OpenAI leaked key notification on April 10, raising the possibility that secrets had already been detected externally before disclosure

What matters about this incident is that it was not a single account takeover, but a supply-chain-type incident that led from OAuth trust relationship → internal systems → customer secret exposure. Vercel subsequently improved its environment variable overview and sensitive variable management UI, but since the default protection method itself was not changed, defaults and governance remain a point of contention.