StreamZilla AI training and your videos

AI training and your videos: what you need to know

Faces, voices, places and knowledge. Once they are in a model, they stay there.

AI models are trained on enormous amounts of text, images, audio and video, much of it collected from the internet. Videos are especially valuable for that: they show real people, real voices and real places, in motion.

For organisations that publish video, that raises questions. Who may use your videos to train AI? What happens to the people who appear in them? And what can you do about it? This guide explains it in plain language.

Join the free private beta

Free during the beta. For organisations. Nothing to pay.

How videos end up in training sets

Crawling. Public web pages, including watch pages, embeds and video files, are collected automatically by crawlers that gather training data.
Platform terms. Some platforms reserve the right in their terms to use uploaded content to improve or develop their services, which can include AI. Check the terms of every provider you use.
AI services in the chain. When subtitles, transcripts or translations are made by an external AI service, your audio and video pass through that service. Its terms decide what happens next.

What is at stake

Faces and voices. Employees, patients, students, children, customers: anyone in your videos can be recognised, and their face or voice can be used to train recognition systems, or to imitate them.
Locations. Buildings, street views, signs, landmarks and even the view from a window reveal where people live, work or study.
What is in the background. Screens, documents, whiteboards and name badges can expose information that was never meant to be shared.
Your knowledge. Lectures, training material and expertise you built up can be used to train models that compete with you, without credit or payment.
No way back. Once a video has been used for training, it cannot be taken out of the model again.

What European law says

GDPR. Images and voices of recognisable people are personal data. Biometric data used to identify someone is a special category with extra protection. Anyone who processes it needs a legal basis, and so do you when you publish it.
AI Act. Since February 2025, the EU AI Act prohibits creating or expanding facial recognition databases through untargeted scraping of facial images from the internet or CCTV footage.
Copyright. EU copyright law allows text and data mining, but rights holders may reserve their rights in a machine-readable way. Providers of general-purpose AI models must have a policy to respect such reservations.

This guide is general information, not legal advice. Ask your privacy officer or lawyer about your specific situation.

What you can do

Check your provider’s terms for any right to use your content for AI or service development.
Reserve your rights with a machine-readable do-not-train signal on your pages and files.
Limit access to videos that are not meant for everyone: unlisted or password-protected.
Look at the picture: who is recognisable, what is in the background, and do you have permission?
Choose AI that stays home: subtitles and transcripts made on European infrastructure, without sending your media to an external AI service.

StreamZilla does all of this by default. See no AI training on your videos and private video hosting.

Frequently asked questions

Can AI companies legally train on my public videos?

It depends on the content and on the rules that apply. Personal data in your videos is protected by the GDPR, the AI Act forbids untargeted scraping of facial images for recognition databases, and you can reserve your copyright with a machine-readable signal. This is general information, not legal advice.

Does a do-not-train signal work?

It is the machine-readable way European copyright law recognises to reserve your rights, and providers of general-purpose AI models must have a policy to respect it. Combined with access restrictions, it gives you real control.

What about AI subtitles?

Check where they are made. If an external AI service processes your audio, its terms apply. StreamZilla makes subtitles and transcripts on its own European infrastructure.

Try Europe’s video platform.

StreamZilla is the all-in-one video platform from the team that put the first live stream on the web, with 32 years of experience. It is in private beta, and joining is free: you get early access, a direct line to the people who build it, and a say in what comes next.