Vercel AI Gateway Does Not Use Data for Training Regardless of Plan
Vercel AI Gateway is an extremely useful service that allows you to easily switch between multiple LLM providers and log and analyze requests. However, one of the biggest concerns when using AI-related services is the privacy issue of “how is the data I input handled?”
As of September 4, 2025, Vercel’s terms of service (AI Product Terms) did not clearly specify the data usage policy for non-Enterprise plan users, especially when using AI Gateway in BYOK (Bring Your Own Key) format. This ambiguity was a concern for some developers.
In response to this, I posted a question about this point in Vercel’s official community and received a response from the Vercel team.
To conclude, Vercel does not use any data passing through AI Gateway for training their own AI products or services, regardless of the customer’s plan or usage method (whether BYOK or not).
Below is the original response from the Vercel team and a summary.
Source: Clarify BYOK Usage in AI Gateway Terms - Feedback - Vercel Community
The team will update documentation for clarity. The updated document isn’t ready yet, but they let me know that:
Vercel does not train on any data for AI Gateway for any customers of any tier for BYOK or otherwise. Vercel does not have any plans to use customer content from AI Gateway for any training unless the customer opts-in. As to third-party LLM providers, it is the customer’s responsibility to review and understand the terms for the providers that they select, which are listed in our Notices and License Information (NLI).
Vercel does not use AI Gateway data from any customers (all plans, BYOK or otherwise) for training. Vercel has no plans to use customer content from AI Gateway for training unless customers opt in.
Regarding third-party LLM providers, it is the customer’s responsibility to review and understand the terms of service for the providers they select.
This response clarified the following points:
However, an important caveat is that the data policies of LLM providers beyond AI Gateway (such as OpenAI or Anthropic) are a separate matter. AI Gateway is merely an intermediary, and how the providers that ultimately receive requests handle that data follows their respective terms of service. Developers need to review the terms of the LLM providers they use themselves.
This clear response from Vercel will be a great source of reassurance for many developers. With privacy concerns about data now addressed, we can utilize the powerful features of Vercel AI Gateway with greater confidence.
Vercel plans to update their documentation to reflect this content in the future. The company’s transparent approach is very commendable. Now we can use AI Gateway with peace of mind!
Above was our investigation into Vercel AI Gateway’s data usage policy.
That’s all from the Gemba.