My Vercel account just got hijacked — what do I do?

Updated

April 2026 Vercel × Context.ai showed: 'Allow All' on a third-party AI tool = account compromise when the tool gets breached. Here's the response playbook.

You wake up to a notification: 'Unusual activity on your Vercel account.' You log in — there's a new deployment you didn't make. You check Settings → Authorized Apps. There are two apps you don't recognize, both granted Allow All.

What happens next

  1. Minute 0 — revoke

    Vercel Dashboard → Account Settings → Authorized Apps → Revoke every app you don't recognize.

  2. Minute 0-5 — rotate

    Rotate every credential the affected apps could have stored: Vercel API tokens, deployment hooks, environment variables.

  3. Minute 5-15 — audit

    Check audit log: every deployment, every env-var change, every team-member add in the past 30 days. Anything you didn't do.

  4. Minute 15-60 — contain

    Force-revoke all sessions (Vercel Dashboard → Settings → Sessions → Sign out everywhere). Re-MFA-enroll. Re-create any leaked Vercel API tokens.

  5. Hour 0-72 — customer impact

    If the hijacker deployed code, audit what's running. If they pulled env vars, every credential in env vars is potentially exfiltrated.

Without Securie

You manually audit OAuth-apps quarterly. You hope you catch them before they get breached. The Apr 2026 wave shows OAuth-app overreach is now a routine attacker objective.

With Securie

Securie's mcp-guard crate's TrustedCatalog + Validator + ScopeGuard layers reject any third-party tool whose declared scope drifts wider than the operator-pinned baseline. The customer-side fix is OAuth-app review hygiene + default-narrow scopes; Securie's role is detecting scope drift after install.

Exactly what to do right now

  1. Revoke every authorized app you don't recognize at Vercel Dashboard → Authorized Apps
  2. Rotate every Vercel API token + every deployment hook + every env var the apps could have stored
  3. Force-revoke all sessions; re-MFA-enroll
  4. Read /incidents/vercel-context-ai-breach-2026 for the canonical case
  5. Quarterly OAuth-app review on every SaaS account (calendar reminder)
  6. Default-narrow scopes on every new app — never grant 'Allow All'