Vercel Breach Highlights Risks of Shadow AI and OAuth Sprawl

· Original article ↗

Summary

The sponsored article argues that employee-approved OAuth links to AI and other SaaS apps can expose enterprise accounts, using Vercel’s fallout from Context.ai as an example, and recommends restricting consent and auditing integrations across SaaS.

Key points

  • A Vercel employee connected Context.ai’s deprecated AI Office Suite to Google Workspace through OAuth, although Vercel was not a Context.ai customer.
  • The article says an attacker used OAuth tokens in Context.ai’s environment to access downstream accounts, including the employee’s Google Workspace; an infostealer infection was alleged as the cause of Context.ai’s compromise.
  • The affected Vercel account had access to internal dashboards, employee records, API keys, NPM tokens, and GitHub tokens.
  • The article describes shadow apps, tenants, browser extensions, and OAuth integrations as risks, and argues that AI tools can expand SaaS interconnectivity.
  • It cites 2025 OAuth supply-chain attacks involving Salesloft Drift and Gainsight, reporting more than 1,000 impacted organizations and over 1.5 billion records stolen.
  • The article reports a 37-fold increase in device-code phishing attacks this year and says more than a dozen criminal phishing-as-a-service kits are in circulation.
  • Recommended measures include default-deny OAuth consent, routine audits of existing integrations, and visibility into OAuth grants beyond Google and Microsoft.

Article Details

Defense Focus
Detecting and reducing OAuth integration sprawl across enterprise and SaaS applications, including browser-based OAuth requests and attacks.
Detection Methods
  • Routinely audit existing OAuth integrations, including their permissions and user counts, and verify that each is still required.
  • Monitor employee app logins and the authentication details described in the article, including MFA use, password security, and SSO.
  • Observe OAuth connection requests as they pass through the browser; Push Security says its platform can detect and block these requests.
  • Analyze browser pages, sessions, and tabs for browser-based threats in real time, as Push Security describes.
Data Sources
  • OAuth grants and integration metadata, including permissions and user counts.
  • Browser-observed app logins and authentication details, including MFA, password security, and SSO.
  • Web pages and browser sessions and tabs where OAuth requests and browser-based attacks may occur.
Platforms
  • Browser-based security platform
Defensive Actions
  • Use a default-deny policy for user consent to new integrations in primary enterprise applications, requiring admin approval.
  • Regularly audit existing OAuth grants and remove integrations that are no longer needed.
  • Extend OAuth visibility and controls beyond primary enterprise cloud applications to SaaS-to-SaaS connections.
  • Maintain an up-to-date inventory of applications and review OAuth grants across them.
  • Detect and block unauthorized OAuth connection requests in the browser.
  • Use browser-based controls to detect and block the attack types described in the article, including device code phishing, malicious extensions, AiTM phishing, credential stuffing, ClickFix, and session hijacking.

MITRE ATT&CK

Threat Actors

Vendors

Products

Related Articles