Case vercel-context-aiAI supply chainUnited States
Compromised Context.ai AI tool's OAuth grant leads to Vercel breach
Attackers who compromised the Context.ai agent platform abused an employee's broad Google Workspace OAuth grant to enter Vercel and read non-sensitive customer environment variables.
HighAttack on AIStatus: Under investigationVerified case file
Open in the interactive mapCite this case
- Reported
- Apr 19, 2026
- Location
- United States
- Target
- Vercel / Context.ai
- Severity
- High
- Triggered by
- Unknown threat actor; data-sale claims on BreachForums unverified
- Actor type
- Unknown
- Reported loss
- None reported
- CVE
- None
What happened
Attackers who compromised the Context.ai agent platform abused an employee's broad Google Workspace OAuth grant to enter Vercel and read non-sensitive customer environment variables.
Impact: Attacker accessed some Vercel environments and customer environment variables not marked sensitive for a limited set of customers; sensitive-flagged variables showed no evidence of access
Who triggered it
Actor: Unknown threat actor; data-sale claims on BreachForums unverified
Attribution: No public attribution; Vercel self-disclosed and traced origin to Context.ai compromise with Google Mandiant
Motive: Credential and data theft; possible resale (unverified extortion claims)
How it happened
Attack vector: Compromised Context.ai Google Workspace OAuth application used to take over a Vercel employee's corporate Google account, then pivot into Vercel environments
Root cause: Broad OAuth scopes granted to a third-party AI tool, and Vercel environment variables not marked sensitive being readable in plaintext
Timeline
- Context.ai employee reportedly infected with Lumma Stealer; OAuth tokens later compromised
- Context.ai discloses unauthorized access to its AWS environment and hires CrowdStrike
- Vercel detects unauthorized access to internal systems via employee Google Workspace account
- Vercel publishes bulletin, notifies affected customers and publishes OAuth app IOC
Response
Vercel engaged Google Mandiant, notified affected customers, recommended credential rotation, published an OAuth app IOC and confirmed with GitHub, npm and Socket that its packages were untouched
Current status: Under investigation
Sources (open in a new tab)
- Trend Micro trendmicro.com
- Help Net Security helpnetsecurity.com
- The Register theregister.com
Cite this case
Compromised Context.ai AI tool's OAuth grant leads to Vercel breach. AI Threat Watch, case vercel-context-ai, Apr 19, 2026. https://aithreat.io/cases/vercel-context-ai/
Permanent link: https://aithreat.io/cases/vercel-context-ai/