In September we wrote about two AI problems Long Play never had: low-effort AI content flooding the feeds, and platforms quietly training AI on the work you upload. This is part two, because the second problem has moved from a worry to the courts.
This month
On 2 November a federal judge in New York refused to throw out a class action accusing Runway, an AI video company, of scraping videos people had posted to YouTube to train its model (Law360). The case goes ahead.
A day later, Matti Klemm - the German voice of Jason Momoa and Mahershala Ali - said he plans to sue Netflix. He says data Netflix disclosed to him showed his voice recordings had been used to train AI systems without his knowledge or consent, and without payment. Netflix says its use of AI follows the law (Yahoo).
And a class action filed in August is still working its way through the courts in San Francisco. A Twitch streamer alleges that Twitch gave creators' streams and videos to Amazon to train AI (FindLaw).
These are allegations, not findings. None of these cases have been decided.
The setting you have to find
The Twitch case shows the pattern most clearly. In August Twitch announced that Amazon would use creator content to train AI, with a setting called "Training for Generative AI" that you had to switch off if you didn't want to take part. According to the complaint, Twitch's chief product officer defended the opt-out approach by saying almost no one would opt in.
That one sentence is the whole problem. If almost nobody would say yes when asked, switching it on for everyone isn't consent. It's a decision made for you, with an exit you have to know exists.
Opt-out works because most people never see the setting. You're busy making things. Defaults are where a platform decides what "normal" is, and an opt-out default for AI training quietly decides that your work is raw material unless you object.
Why it keeps happening
Video is expensive to make and valuable to train on. Any platform holding a big library of it is sitting on something AI companies want. Terms written years ago, granting broad rights to "operate and improve our services", get stretched to cover uses nobody had in mind when they clicked agree. The Twitch complaint argues training may have started before the setting was even announced.
Our answer
Long Play doesn't have that setting, on or off, because there's nothing to switch. We will never use your videos to train AI, and we will never license them to anyone who does. That's on our privacy page and in our pledge, and it isn't changing.
Where Long Play does use this kind of technology, it works for you and stays here. If you ask for automatic captions, speech recognition runs on Long Play's own servers with an open-source model. Your audio isn't sent to an outside AI company and isn't used to train anything. The captions are made for your video, and that's the end of it.
What you can do
Wherever your work lives, it's worth a few minutes to check:
- Search the platform's settings for "AI", "training" or "machine learning", and see what the default is.
- Read the content licence in the terms, not just the summary. Watch for broad phrases like "improve our services" or "develop new products".
- Keep your original files somewhere you control.
And if you'd rather not have to keep checking, that's what Long Play's principles are for.