Twitch is facing criticism from creators and users after revealing that content shared on the streaming platform can be used to train generative artificial intelligence models developed by its parent company, Amazon.
The controversy centres on a setting that allows Amazon to use content generated by Twitch creators and their audiences for AI training. The option is enabled by default, meaning users must manually switch it off if they do not want their content included.
Twitch has confirmed that users can opt out of the programme, but the decision to make participation the default has triggered strong reactions from parts of the streaming community.
During a livestream discussing the changes, Twitch chief product officer Mike Minton acknowledged that significantly fewer people would participate if users were required to actively opt in.
The company says content from users who remain opted in could be used to improve Amazon’s generative AI models. Twitch has said the information will not be resold to other companies for AI training.
The types of content that could potentially be used include livestreams, clips, images, audio, chat messages and other material associated with a creator’s channel.
One potential application involves speech recognition. Audio from streams could help improve AI systems that convert speech into text, potentially improving automatic captions on Twitch and other Amazon services.
However, the announcement has also raised questions about copyrighted material appearing within livestreams.
Gaming represents a significant portion of Twitch’s content, meaning streams regularly contain video game graphics, dialogue, music and other intellectual property belonging to game developers and publishers. Some users have questioned how such material could be handled when training AI systems.
The controversy reflects a wider debate across the technology industry about how companies obtain the enormous quantities of data required to train increasingly powerful generative AI models.
AI developers require large datasets containing text, images, audio, video and other information to train systems capable of generating new content. As competition in artificial intelligence intensifies, technology companies are increasingly looking at content available across their own platforms as a valuable source of training data.
For creators who do not want their Twitch content used for generative AI training, the platform provides an option within account settings.
Users can open the Streamer Dashboard, navigate to Settings, select Security and Privacy, and locate the option relating to generative AI training. Turning the setting off is intended to prevent eligible channel content from being used for Amazon’s generative AI models.
Some Twitch users have reported difficulties with the setting, claiming it appeared to switch back on after they disabled it. It is unclear how widespread the issue is.
The decision to use an opt-out rather than opt-in system has become one of the biggest points of criticism.
Some creators argue that users should have to explicitly give permission before their work is included in AI training datasets, rather than being automatically included and required to manually withdraw.
The dispute also raises broader questions about consent and ownership as generative AI becomes increasingly integrated into major online platforms.
Social networks, video platforms, search engines and other digital services collectively hold enormous amounts of user-generated information that could potentially be valuable for developing artificial intelligence systems.
For technology companies, this data could help create better AI products. For creators, however, the concern is whether their work should contribute to commercial AI systems without explicit permission or compensation.
Amazon acquired Twitch in 2014 for approximately $1 billion and has since become one of the world’s biggest investors in artificial intelligence and cloud computing.
The company is competing with technology giants including Google, Microsoft and Meta as businesses invest heavily in generative AI infrastructure, models and applications.
As AI becomes more deeply integrated into online platforms, the Twitch controversy demonstrates that one of the industry’s biggest challenges may not simply be building more capable AI.
It will also be deciding who controls the data used to train it — and whether users should have to opt in rather than opt out.
