{"slug": "twitch-admits-it-made-ai-training-opt-out-because-no-one-would-opt-in", "title": "Twitch Admits It Made AI Training Opt-Out Because No One Would Opt In", "summary": "Twitch, owned by Amazon, admitted on August 12, 2026, that it had automatically opted all users into training Amazon's generative AI models on their content, with an opt-out toggle added only after creator backlash. Chief product officer Mike Minton said on a livestream that an opt-in approach would have meant 'nobody would opt in,' while head of community Mary Kish noted the toggle exists due to community pressure. The setting, which requires manual discovery, does not protect users' chat messages in other channels, and Twitch has not reversed the default.", "body_md": "*Twitch quietly opted every user into training Amazon's AI on their content. Only after creators noticed did it add a way to say no - and its own product chief admitted, on camera, that an honest opt-in would have persuaded almost nobody to sign up.*\n\nOn August 12, 2026, Twitch added a \"Training for Generative AI\" toggle under Settings, Security and Privacy. Turn it on and you stop your streams, VODs, clips, chat logs and channel page content training Amazon's generative AI models. Here's the catch. The setting had already been switched on for every existing account before the toggle even existed. Millions of creators had been feeding Amazon's models by default, with no idea a switch existed to stop it, according to reporting from TechCrunch and Engadget.\n\nTwitch didn't send an email. There was no pop-up when you logged in. The company posted the change through its TwitchSupport account on X, linked to a help article, and left it there. The story only spread because streaming personality Zach Bussey posted a screenshot of the buried setting, and other creators passed it around from there.\n\nDays later, at a Q&A on Twitch's own channel, someone asked chief product officer Mike Minton why the setting wasn't opt-in from the start. His answer, reported by PC Gamer and VGC, was blunt: \"If it was opt-in, nobody would opt in.\" Twitch's head of community, Mary Kish, said the toggle exists at all because of community pushback against AI training on creator content. So the company built the escape hatch. Then it hid the door.\n\nHere's the part that makes the whole thing messier. Post in someone else's chat, and that message is governed by the channel owner's AI-training setting, not yours. Turn off training on your own channel, and your chat messages elsewhere are still fair game - if the streamer you're watching hasn't flipped the same switch. A creator who carefully opts out can still end up training Amazon's models every time they type in someone else's stream.\n\n[Amazon's Zoox Wins First US Approval to Charge Fares in a Robotaxi With No Steering Wheel](https://startupfortune.com/amazons-zoox-wins-first-us-approval-to-charge-fares-in-a-robotaxi-with-no-steering-wheel/)\n\nNHTSA gave Amazon-owned Zoox the first US federal approval to charge passengers for rides in a purpose-built, steering-wheel-free, pedal-free robotaxi. Zoox starts charging fares in Las Vegas on August 10, 2026, after months of free rides, under a two-year exemption capped at 2,500 vehicles a year.\n\nKotaku ran a follow-up on the backlash, documenting streamers venting about a policy that assumed consent instead of asking for it. The frustration isn't really about the setting itself. It's about the sequence: switch it on for everyone, collect months of training data, then quietly hand out an opt-out once people notice.\n\nTwitch isn't inventing this playbook. It's just being unusually candid about it. Reddit signed a content-licensing deal with Google worth roughly $60 million a year back in 2024, feeding its forums straight into AI training. X rewrote its privacy policy that same year to let Grok and other \"collaborators\" train on user posts unless people dug through settings to say no. YouTube has faced its own creator pressure over AI training on video libraries. What's different here is simple: a senior Twitch executive said the quiet part out loud, on a livestream, to the very creators whose content is on the line.\n\nEvery platform running an opt-out default is making the same bet Minton described out loud. Admitting it on camera doesn't make the practice less extractive. It just removes the pretence. As of this week, Twitch hasn't reversed the default or promised to make the setting opt-in. It still starts on - unless a creator finds it themselves.\n\n**Also read:** [Wyoming Woman Sues xAI Over Grok Generated Explicit Images of Her as a Child](https://startupfortune.com/wyoming-woman-sues-xai-over-grok-generated-explicit-images-of-her-as-a-child/) • [Unitree's Robot IPO Just Became Shanghai's Most Oversubscribed Tech Listing Ever](https://startupfortune.com/unitrees-robot-ipo-just-became-shanghais-most-oversubscribed-tech-listing-ever/) • [LG Bets Its Actuators and Batteries on a Humanoid Robot Built With Nvidia](https://startupfortune.com/lg-bets-its-actuators-and-batteries-on-a-humanoid-robot-built-with-nvidia/)", "url": "https://wpnews.pro/news/twitch-admits-it-made-ai-training-opt-out-because-no-one-would-opt-in", "canonical_source": "https://startupfortune.com/twitch-admits-it-made-ai-training-opt-out-because-no-one-would-opt-in/", "published_at": "2026-08-16 07:02:12+00:00", "updated_at": "2026-08-16 07:11:20.269650+00:00", "lang": "en", "topics": ["generative-ai", "ai-policy", "ai-ethics"], "entities": ["Twitch", "Amazon", "Mike Minton", "Mary Kish", "Zach Bussey", "TechCrunch", "Engadget", "PC Gamer"], "alternates": {"html": "https://wpnews.pro/news/twitch-admits-it-made-ai-training-opt-out-because-no-one-would-opt-in", "markdown": "https://wpnews.pro/news/twitch-admits-it-made-ai-training-opt-out-because-no-one-would-opt-in.md", "text": "https://wpnews.pro/news/twitch-admits-it-made-ai-training-opt-out-because-no-one-would-opt-in.txt", "jsonld": "https://wpnews.pro/news/twitch-admits-it-made-ai-training-opt-out-because-no-one-would-opt-in.jsonld"}}