Vercel Breached After Third-Party AI Tool Compromise

Cloud hosting giant Vercel has confirmed a significant security breach resulting from a compromise of Context.ai, a third-party AI tool. The incident allowed sophisticated attackers to infiltrate Vercel’s internal systems and access the credentials of a limited subset of customers. According to a security statement released Sunday, the breach began when an attacker exploited a compromise of Context.ai to take over a Vercel employee’s Google Workspace account. This unauthorized access enabled the threat actors to enumerate Vercel environments and extract variables that were not specifically designated as sensitive.

Vercel CEO Guillermo Rauch noted that while customer environment variables are encrypted at rest, the attackers managed to bypass various defense-in-depth mechanisms through rapid enumeration. Rauch characterized the attacking group as highly sophisticated and likely accelerated by artificial intelligence, noting the surprising velocity and depth of their understanding of Vercel’s infrastructure. In response, Vercel has notified all affected customers directly, advising them to immediately rotate their credentials, environment variables, and Deployment Protection tokens. The company has also engaged Google’s Mandiant team and law enforcement to assist in a comprehensive forensic investigation.

The origin of the incident has been traced back to a Google Workspace OAuth app associated with Context.ai. Cybersecurity researchers at Hudson Rock reported that the initial entry point may have been a Lumma stealer infection on a Context.ai employee's machine in February 2026. The infection allegedly occurred after the employee attempted to download game exploits, leading to the theft of high-level corporate credentials, including access to administrative accounts. These stolen credentials provided the leverage necessary for the attackers to escalate privileges and eventually pivot into Vercel’s enterprise environment.

While a post on BreachForums briefly claimed that the prolific hacking group ShinyHunters was responsible for the theft, the group has since denied involvement, and the post was removed. Context.ai has issued its own advisory confirming unauthorized access to its AWS environment and noting that at least one Vercel employee had granted "Allow All" permissions to the AI suite using an enterprise account. Vercel has since deployed additional monitoring and protection measures, confirming that its core open-source projects, including Next.js and Turbopack, remains secure and unaffected by the breach.