Dutch regulator urges Twitch users to opt out of Amazon AI training

This regulatory advisory highlights the growing tension between streaming platforms seeking proprietary training data and European privacy standards. By utilizing live broadcast content for machine learning, Amazon risks significant pushback from regulators who view biometric data in streams as highly sensitive. For the broader streaming ecosystem, this signals a shift where default data-sharing policies for AI development will face intense scrutiny, potentially forcing platforms to adopt explicit opt-in models. The immediate impact could be a reduction in available training data if users follow the AP's guidance. Watch for whether other EU data protection authorities follow the Dutch lead to open formal investigations into Twitch's default consent mechanisms.

The Dutch AP's advisory to Twitch users arrives amid a broader European regulatory push against default opt-in mechanisms for AI training data. In April 2025, Meta announced it would resume training its AI models on public content from EU users after pausing the effort in mid-2024 under pressure from the Irish Data Protection Commission. The DPC had intervened after advocacy group NOYB filed complaints across multiple European data protection authorities. Meta's resumption followed a December 2024 opinion from the European Data Protection Board affirming that its original approach met legal obligations, though the company agreed to enhanced safeguards on transparency, objection rights, and de-identification. The Twitch situation differs critically because Amazon's default setting is opt-out rather than opt-in, a distinction that places it on weaker legal ground than Meta's revised framework.

Amazon updated Twitch's account settings to include an AI training opt-out toggle, but WIRED reported that the default remains enabled, meaning streamers must actively disable the setting to prevent their content from being used. This opt-out architecture contrasts with the consent model Meta adopted after its DPC engagement, which included in-app and email notifications with a direct link to an objection form. The AP's deputy chair Monique Verdier specifically flagged that facial images and voice recordings captured during live streams constitute biometric data under Article 9 of the GDPR, which demands explicit consent rather than passive opt-out defaults. In May 2025, the Irish DPC formally approved Meta's revised EU/EEA LLM training plans after the company implemented enhanced GDPR safeguards including risk assessments and de-identification measures, establishing a compliance benchmark that Twitch's current approach does not appear to meet.

The competitive and technical implications extend across the streaming and creator economy. Meta explicitly stated it was following the example set by Google and OpenAI, both of which had already used data from European users to train their AI models, signaling that large-scale training on user-generated content is becoming an industry norm that regulators must address systematically. Meanwhile, the DPC announced it was investigating xAI's training of Grok, indicating that enforcement attention is expanding beyond a single company. For Twitch, the Dutch advisory establishes a regulatory precedent that other EU member-state authorities could adopt, potentially forcing and restructure its consent mechanism from opt-out to explicit opt-in. The AP's emphasis on biometric data in live streams, including real-time facial expressions and voice patterns, raises the stakes beyond text-based content because anonymization of such multimodal data without destroying training utility remains technically unresolved.