Cloud platform company Vercel disclosed a security incident in April 2026 that began when an employee granted an AI productivity tool, Context.ai, blanket OAuth permissions to a corporate Google Workspace account. The tool was authorized with an Allow All scope, giving it wide access to the employee's workspace data.
Attackers who had separately compromised a Context.ai employee used the resulting OAuth tokens to take over the Vercel employee's account. From there, they moved laterally into Vercel's internal systems, according to the company's disclosure of the event.
The incident illustrates a category of risk tied to AI assistants and automation tools that request broad, persistent access to corporate accounts. Once granted, those tokens can become a pathway for attackers if the third-party AI vendor is itself breached, bypassing some traditional account protections.
Security analysts reviewing the event pointed to the practice of approving wide-ranging permissions for AI tools without scoping access to only what a task requires. Vercel said it responded to contain the intrusion after detecting the unauthorized access tied to the compromised tokens.
Source: BlueRadius - https://blueradius.io/ai-cybersecurity-incident-report-2026