Vercel just updated its Terms of Service to allow the company to train AI models on code deployed by Hobby (free tier) and Trial Pro users. The change is opt-out, not opt-in, meaning your code is included unless you manually disable it before March 31, 2026.
The data Vercel plans to use includes code and Vercel agent chats, build and deployment telemetry (including build errors), and aggregate traffic stats. The company says it strips personal information, account details, environment variables, and API keys using "industry-standard" anonymization before anything enters a training pipeline.
Who's Affected and Who Isn't
Paid Pro and Enterprise users default to opted out. Only Hobby plan users and Trial Pro accounts are enrolled automatically. Open-source projects like Next.js and shadcn/ui are explicitly excluded from the training data pool.
To opt out, Hobby users need to go to Team Settings → Data Preferences before the March 31 deadline. You can still opt out after that date, but it only applies going forward. Any data already collected before your opt-out can't be retroactively removed.
The Quiet Part
This follows a familiar playbook: offer a free tier, build a massive user base, then change the terms to monetize the data those users generated. Vercel frames this around "agentic infrastructure capabilities" and improving autonomous features, but the practical effect is that hundreds of thousands of side projects, portfolio sites, and hobby apps just became training fodder.
The 12-day window between announcement (March 19) and deadline (March 31) is tight. Most Hobby plan users don't read changelogs. The people most likely to notice and opt out are the same developers who'd be most uncomfortable with their code being used this way.
If you're on Vercel's free tier and don't want your code in their AI training data, go to Team Settings now. The opt-out takes about 30 seconds. The re-training of a model on your code is permanent.