«If it were voluntary, no one would agree»: Twitch starts training AI on user content by default

Streaming platform Twitch, owned by tech giant Amazon, has become the target of widespread community dissatisfaction following the introduction of a new data usage policy. The service has begun using video content from user channels to train its own generative artificial intelligence models. The key point that sparked outrage is that this feature was enabled by default, forcing users to manually search for settings to opt out of the project.

Scandal over «silent consent» and community reaction

Information about the ability to opt out of using content for training neural networks only became known after users noticed changes in the terms of service and spotted relevant clauses in the settings. The announcement triggered sharp criticism from streamers and viewers. Many content creators believe the platform should have requested active consent (opt-in) rather than forcing users to cancel the mandatory activation of the feature (opt-out). Concerns arise because stream materials, often containing original scripts and unique styles, can now be used by Amazon for commercial purposes without direct financial compensation to the creators.

Justification from Twitch leadership

During a Patch Notes stream dedicated to platform updates, Twitch’s Chief Product Officer Mike Minton explained the company’s logic. According to him, if the feature were enabled on a voluntary basis, «no one would agree». Minton stated that their approach is unique because they provide the option to opt out of model training, which, in his view, demonstrates respect for user choice. He also noted that enabling such features by default is standard for most content services worldwide. Thus, Twitch leadership attempted to legitimize their actions by citing industry norms and the low likelihood of voluntary user participation.

Technical nuances and exceptions

It is important to understand that even disabling the «AI Training» parameter does not give users full protection. Twitch and Amazon administration reserve the right to use channel content for other purposes. Specifically, stream materials can still be processed by algorithms to improve platform features such as viewer search, content recommendations, and community safety via the AutoMod system. This means that full control over one’s digital footprint on the platform remains questionable, and users only limit the use of their data for training external Amazon generative models.

Contradictory data

There is a discrepancy between Twitch leadership’s position and community expectations. On one hand, Mike Minton claims that providing an opt-out option is an act of respect for users. On the other hand, experts and users point out that the phrasing «no one would agree» sounds like an admission that the platform cannot offer users anything valuable in exchange for their data, and is therefore forced to use a mandatory model. Furthermore, although some sources mention that Twitch tends to reverse unpopular decisions, in this case Amazon has strategically tied its future to AI development, making the likelihood of quickly canceling this feature extremely low despite the criticism.

Outlook and conclusions

The situation with Twitch and AI training on user content reflects a global trend in the tech industry of 2026. Companies are increasingly using massive amounts of user data to develop their neural networks, often bypassing ethical norms. Users wishing to protect their content must carefully review privacy settings on the platform and manually disable participation in model training. However, as practice shows, such measures only partially solve the problem, leaving corporations with broad opportunities to use data within the framework of internal service optimization.