Twitch is now feeding creator content to Amazon’s generative AI models by default, and the company isn’t pretending otherwise about why. During a livestream addressing nearly 3,000 angry users, Chief Product Officer Mike Minton gave a blunt answer when asked why the program isn’t opt-in: “If this was opt-in, nobody would opt in. That’s honestly the answer.”
The admission cuts through the corporate framing Twitch used when announcing the change, which positioned it as “[adding] a setting that lets you opt out” rather than revealing that Amazon would start training on streamer content automatically. For a platform built on livestreamers broadcasting their faces and voices for hours weekly, the stakes are personal. Thousands of hours of audio and video now become training data unless creators dig through channel settings to disable it.
Minton’s candor didn’t extend to clarifying what Amazon has already done. When a user asked whether their videos had been used for training prior to this announcement, Minton said he didn’t actually know what Amazon had used for model training. That gap between Twitch’s messaging and Amazon’s data practices is precisely what has streamers worried they’ve been training models without consent or knowledge.
Twitch Head of Community Mary Kish tried to soften the blow by pointing to Meta, which already trains on public Facebook and Instagram content. But that comparison lands poorly — Meta’s approach has faced regulatory pressure in the U.K., where users can opt out, while most global users have no meaningful recourse. Twitch framing opt-out as a community concession rather than a baseline right doesn’t change the underlying dynamic: the default benefits Amazon, not creators whose livelihoods depend on the platform. The real question now is whether streamers will accept this quietly or follow the pattern of past creator revolts that forced platform policy reversals.
