On April 19, Vercel confirmed what nobody wants to read: someone gained unauthorized access to "certain internal Vercel systems." The company's security bulletin is deliberately vague. Incident response teams are involved. Law enforcement has been notified of a security breach.
Only a "limited subset of customers" were directly impacted, Vercel claims. Those customers are being contacted individually. Everyone else is left reading between the lines of a thin advisory.
The advisory tells all customers to review their environment variables and turn on the "sensitive environment variable" feature. Sensible advice. But it's left users on Hacker News guessing about actual exposure. One commenter running production workloads still doesn't know if their data leaked. Others question whether the sensitive flag even protects variables from internal access.
Consider the timing. Vercel has been expanding into AI agents. The Vercel Agent, described as "an agent that knows your stack," needs deep access to codebases, deployment configurations, and environment variables. v0, their AI app generator, has similar requirements. So does the AI Gateway for routing model requests. Each product needs more access to do its job. And each product increases the blast radius if something goes wrong. Recent privacy concerns have highlighted the risks.
A single compromised agent system could expose credentials across dozens of customer projects at once.
More updates are coming as the investigation continues. If you deploy on Vercel, rotate your secrets now and enable those sensitive environment variable protections. Don't wait for a notification that may never arrive.