Twitch turned on AI training by default. Its product chief explained why: "nobody would opt in"
A streamer in Connecticut is suing Twitch and Amazon for using his broadcasts to train AI. The most damaging line in the story wasn't written by his lawyers — it was said on Twitch's own show.
Warren Pandiscia streams on Twitch. This week he filed a class action lawsuit against Twitch and its owner Amazon, saying they used his streams and videos as what the complaint calls a "dataset necessary to fuel Amazon's AI products" — without a license and without asking. Courthouse News reported the filing first.
Earlier this month, Twitch Support posted on X that it had added a setting letting you opt out of "having your channel content used to train generative AI content models across Amazon." Then, on Twitch's own recurring Patch Notes stream, a viewer asked chief product officer Mike Minton the obvious question: why is this on by default instead of something people choose to join? His answer: "if it was opt-in, nobody would opt in."
That sentence is the whole story. It is a company saying out loud that it knows people don't want this, and that the setting was designed around that knowledge. Opt-out isn't a technical preference. It's a bet that most people never open the settings menu — and that bet is usually right.
Think of it like a gym that renews your membership unless you cancel in writing. Legal, maybe. Popular, never.
Here's the mechanic that makes this bigger than one streamer. Pandiscia isn't claiming his video was stolen off the open internet — he's claiming breach of contract. He agreed to Twitch's terms to put his content on Twitch. He says using that content as "free training stock for separate commercial AI products" was never part of the deal. If a court agrees, the question stops being "is scraping legal" and becomes "what exactly did people sign up for when they hit Accept." Almost every platform on earth has that same box.
He also isn't first. Earlier this year three YouTubers filed a class action against Apple, saying it scraped copyrighted content to train its models. Different company, same shape of argument: the people who made the material found out afterwards.
Anyone who posts to make money or to build something — streamers, video makers, people whose channel is their side income or their whole income. Their work is now raw material for a product they don't own a piece of. It also touches everyone in their twenties who's been told that building an audience online is a real career path: the terms of that career are being rewritten in settings menus, not contracts you negotiate. And it touches you if you've never streamed a second in your life, because the default on your video, your voice note, your photo library is set by the same logic Minton described.
Two things to watch. First, whether Twitch changes the default to opt-in — that would be an admission the setting was the problem. Second, whether the court lets this proceed as a class action rather than one streamer's dispute. If it's certified as a class, every Twitch streamer covered by those terms is suddenly part of it, and Amazon is arguing against a crowd instead of a person. Also worth watching: whether Minton's sentence shows up in the filings. Companies rarely hand plaintiffs a quote that good.
"If it was opt-in, nobody would opt in." Read it again — it's not a defense of the practice, it's a measurement of consent. Somebody at Twitch already ran the numbers on how many people would say yes, and the answer was approximately none. The lawsuit is just asking whether that answer should have mattered.
Sources: Engadget, Aug. 23, 2026, reporting on a class action filed by streamer Warren Pandiscia, first reported by Courthouse News; Twitch Support post on X; Twitch's Patch Notes stream with chief product officer Mike Minton.
Written by THE TELL’s AI newsroom. how we work · corrections