Vercel breach: an AI tool's Google Workspace grant
Vercel traced its April breach to an AI tool an employee had connected to a work Google account. How to see, and limit, what your own staff have granted.
Published · 3 min read

Vercel, a cloud platform that hosts web applications, disclosed a breach on Sunday 19 April, and by the morning of Monday 20 April, Manila time, had traced it to Context.ai, a small third-party AI tool one of its employees used. The attacker used that access to take over the employee's Google Workspace account, then the employee's Vercel account, and from there read environment variables that a subset of customers had not marked as sensitive, the kind that hold API keys, tokens and database credentials. By Context.ai's account, the attacker appears to have reached Vercel's Google Workspace with a compromised OAuth token: a permission an employee had already granted.
One sign-up, and "Allow All"
Context.ai's statement, quoted by Help Net Security on 20 April, explains how. Vercel was not a Context customer. At least one Vercel employee appears to have signed up for its AI office suite with a Vercel enterprise account and granted "Allow All" permissions, and Vercel's own OAuth settings appear to have let that grant through. Context said an attacker had reached its AWS environment the month before and had likely compromised OAuth tokens belonging to some of its consumer users. Vercel's bulletin says the tool's Google Workspace app was part of a broader compromise that could reach hundreds of its users across many organisations, and it published the app's ID so that administrators could check for it.
Later updates found a small number of additional compromised accounts, and on 16 October Vercel said all remediation was complete and closed the incident.
An integration nobody designed
Joining up what an office already uses is part of my work: systems connected to an HR system, a membership database, QuickBooks or Google Workspace, so that nobody types the same record twice. A connection like that should be designed. Someone decides what it may read and change, it asks for no more than that, and the office can see it listed and switch it off. An AI tool that a member of staff connects with one click is an integration too, only nobody designed it. It can keep its access to mail, files and calendars for as long as the grant stands, and the office may not know it exists until a bulletin like Vercel's arrives.
Five checks for an office on Google Workspace
- In the Google Admin console, under Security, then Access and data control, then API controls, open the list of accessed apps. It shows each third-party app, how many staff use it and which services it reaches.
- Set Gmail, Drive and Calendar to restricted, so that only apps you have trusted can reach them. Untrusted apps stop working and their tokens are revoked, so warn staff first.
- For apps you have not reviewed, block the high-risk scopes Google lists, such as reading or sending all mail and full access to Drive.
- Remove any grant nobody can explain, and ask staff to check with you before connecting an AI tool to a work account.
- Keep secrets out of plain settings. Vercel's advice was to rotate every variable not marked sensitive, and to mark secrets as sensitive from then on.
If that list of connected apps turns out longer than anyone expected, the fit check is a good place to sort out what should be joined up properly and what should be switched off.
Sources
- Vercel, Vercel April 2026 security incident, 19 April 2026
- Help Net Security, Vercel breached via compromised third-party AI tool, 20 April 2026
- Google Workspace Admin Help, Control which apps access Google Workspace data
Enrique Lacambra III builds internal business systems from Cainta, Rizal.
Tell me what is broken and I will tell you if it is worth building. A 20-minute call, no charge, and I will say so if you do not need me.
