Twitch users outraged as Amazon uses their content to train AI in opt-out feature
BBC News ● Covered by 3 sources
Twitch says streams, clips, chats and images can train Amazon’s AI unless users opt out. Streamers are furious because the setting is on by default, and some say it flips back on.
Based on reporting by BBC News — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Twitch is facing a backlash after saying that users who do not opt out may have their channel content used to train Amazon’s AI models. The company laid out the feature in its own FAQs and then walked viewers through how to turn it off on a livestream, but that only seemed to make people angrier.
Mary Kish, Twitch’s head of community, said on the stream that data collection for AI training has become standard across the industry. She also told viewers, plainly, that she did not expect a positive reaction. She was right about that part. The chat filled with criticism, including complaints that the setting should never have been enabled by default.
The opt-out is tucked inside the Streamer Dashboard. Users have to open Settings, go to Security and Privacy, scroll near the bottom, and switch off “training for Generative AI.” If they do, Amazon says it will not use their streams, clips, images, chats and other channel content to train its generative AI systems.
Twitch says some of the material could be used to improve things like speech-to-text, including subtitles on Twitch streams and Amazon videos. It also says the data will not be resold to other companies. But some streamers are already saying the toggle does not behave reliably, with at least one user reporting that it turned back on after being switched off.
There is another layer here too: game developers. Because Twitch streams are full of people playing video games, some users are worrying about what Amazon’s models may absorb from that footage as well. Amazon, which bought Twitch for nearly $1bn in 2014, has been pouring money into AI as it tries to keep up with Google, Meta and the rest. What it gets here is not just data. It is a reminder that creators are expected to do the unpaid work of training the machines.
My take — AI-written commentary, not fact-checked reporting
This is exactly the sort of move that makes people distrust AI: default to yes, then hand users a privacy maze and call it control. Twitch could have set a cleaner standard, but tech companies still prefer quiet consent and loud surprise. The industry keeps acting shocked when creators hate being turned into training fuel.
Read more about this at: BBC News
