Is my data used to train AI? Where the answer is written

It depends on the tool, and the answer is written in two places: the license you grant over your content in its terms of service, and the "how we use information" section of its privacy policy. Search both for the word "train". If neither page says the tool trains on your content or promises it doesn't, the pages leave the question open, and you should ask the vendor in writing.

For Sume, this post quotes the Terms of Service (last updated August 10, 2026) and the Privacy Policy (last updated September 13, 2026), read on 2026-09-29. It is not legal advice.

What wording should I search for in the terms?

Training permission is rarely in a heading. It hides inside a list of purposes. Search the terms and privacy policy for these words, then read the whole sentence each one sits in:

• "Train", "training", "machine learning" and "model". A sentence that says your inputs or outputs may be used to train or develop models answers the question with a yes.
• "Improve". A purpose such as "improve the Services" is broad. Unless the page defines it, it neither confirms nor rules out training.
• "Generalized AI models" or "foundation models". A promise not to train these may still allow narrower models, and may cover only one kind of data.
• "Opt out". If training is allowed, check whether the pages describe a setting or an email request to stop it, and whether it applies to past content.
• Your plan or contract. A signed enterprise agreement can differ from the public terms, so read the one you actually sign.

Does Sume use my data to train AI?

Sume's pages don't say it trains AI models on customer content, and they don't say it doesn't. The only explicit training sentence covers the basic profile and email data from Google sign-in. Here is where each relevant sentence sits:

Does "improve the Services" mean training?

The Terms don't define it, and this post won't read a meaning into it. If the difference matters for your content, for example client footage or a person's face and voice, email the address the Terms give for questions, dev@sume.com, and keep the written answer with your vendor records.

Under the Terms, a signed Order Form or other written agreement controls where its commercial terms conflict with the public Terms, so an enterprise buyer can ask for a no-training clause during that negotiation.

What about the AI models a tool calls?

A tool that routes work to outside models sends them your prompts and media, so their use counts too. Sume's Privacy Policy lists those model providers by category, not by name, and the Terms keep its internal model routing private. The Video generation docs add that with model: "sume/auto", the family that ran is never disclosed. So the written answer to get is Sume's own. Zero data retention for AI video APIs covers what Sume stores and who processes it; How to delete my data from an AI tool covers removing it.