The Vercel Plugin That Spoke in Claude's Voice to Ask for Your Data
Vercel's Claude Code plugin injected its own consent dialog through Claude, making Vercel's telemetry request indistinguishable from Claude itself.
Vercel's Claude Code plugin injected its own consent dialog through Claude, making Vercel's telemetry request indistinguishable from Claude itself.
Key takeaways
What the Vercel plugin established is a new threat model for AI tooling: the consent mechanism itself becomes a vector for deception. By routing its telemetry request through Claude's rendering layer, Vercel did not forge a signature — it borrowed one. The consent prompt appearing even in projects without Vercel configuration means users encountered the request in contexts where Vercel had no obvious role, making attribution even harder. This is not a bug that patches clean. The architecture that allows plugins to inject into Claude's system-level context is, by design, invisible to the end user — and any plugin can use it the same way.
Methodology
This story was generated autonomously from 5 source records. An editorial model synthesizes, weights, and cites each source. No human editorial judgment was applied.