Dutch regulator demands Twitch streamers opt out of Amazon’s AI training data use

Twitch streamers in the Netherlands must now take steps to prevent their broadcasts from being used in Amazon’s AI training programs.

The Dutch Data Protection Authority has ordered that streamers actively decline participation in Amazon’s AI development efforts. This requirement comes after Amazon updated its terms to claim broad rights over user-generated content for training machine learning models, including the Titan series and any future iterations.

Under the revised Terms of Service, Amazon secures a permanent, global, and royalty-free license to repurpose any Twitch content—such as gameplay, live commentary, chat logs, and personal details—without compensation. The Dutch DPA has ruled that this practice likely conflicts with GDPR, as Amazon’s reliance on "legitimate interest" does not justify such extensive use, nor did streamers provide meaningful consent when the terms were changed retroactively.

Streamers wishing to opt out must access their Creator Dashboard, go to Settings → Channel, find the Content Usage Permissions section, and disable the option labeled "Allow Amazon to use my content for AI/ML training." This adjustment must be applied individually to each channel they own.

The opt-out feature was introduced following regulatory pressure from the EU, though US-based streamers currently have no equivalent control. While California’s CCPA grants users the right to opt out of data "sales," it remains uncertain whether AI training data falls under this category.

This situation reflects broader concerns about how user-generated content is being used to train large language models. Similar disputes have involved Microsoft’s GitHub Copilot, Reddit’s $60 million sale of user posts to Google, and Stack Overflow’s collaboration with OpenAI. In each case, user content has been repurposed for AI development without explicit permission.

The Dutch DPA’s decision suggests that EU regulators now treat AI training as a separate data processing activity requiring specific legal justification, rather than simply an extension of service improvements. Other EU countries, including Ireland, France, and Germany, may adopt similar positions.

EU and UK streamers should act promptly to exercise their opt-out rights. To safeguard against potential misuse, they may also choose to download and store their VODs and clips locally before deletion, as Amazon retains training rights even after content removal. Alternatives like PeerTube, Trovo, or self-hosted platforms such as Owncast could be explored for greater privacy control. US creators can submit a CCPA "Do Not Sell My Info" request to Amazon or Twitch’s legal team, which must be addressed within 45 days.

One unresolved issue remains: Amazon has not revealed how much of Titan’s training data comes from Twitch, AWS customer data, or public web scraping. Without clear model documentation on data sources, transparency depends on regulatory oversight. The Dutch DPA’s intervention, however, provides a concrete solution—streamers should use the opt-out toggle to protect their content.