
A Twitch streamer sued Amazon and Twitch for using creator videos to train AI without permission or pay. The platform confirmed it uses streams for AI training with an opt-out setting added Aug.
12.
Channels are automatically enrolled, meaning creators must manually disable the feature to prevent their content from being used.
What happened
A Connecticut-based Twitch streamer, Warren Pandiscia, filed a class action lawsuit in the Northern District of California on Thursday against Twitch and parent company Amazon, claiming they harvested millions of creator videos without permission to train Amazon's generative AI models. Twitch confirmed on Aug. 12 that it uses creators' content for AI training and added an opt-out setting, but channels are automatically enrolled unless users manually disable the feature.
Why it matters
Pandiscia argues that because Amazon's AI products are commercialized, the company had financial incentive to acquire training data at scale without negotiating licenses or seeking permission. He claims the practice violates creators' implied contracts with Twitch and state unfair competition law, and that content scraped as early as 2024 cannot be recovered. Twitch's own former Chief Monetization Officer Mike Minton acknowledged in 2024 that Twitch content was already being used for AI development, though he characterized it as prototyping rather than production scale.
What to watch
Pandiscia seeks injunctive relief, damages, restitution, and disgorgement of profits. The lawsuit highlights that Twitch's opt-out approach means content from users who opt out may still be used if they engage on opted-in channels—a design flaw Pandiscia argues makes true consent impossible. The case also raises questions about whether Twitch disclosed the AI training practice and opt-out setting transparently to creators before updating its Terms of Service and Privacy Policy on Aug. 12.
Ask the AI about this article →
The lawsuit centers on a fundamental tension between Twitch's business model and creator rights. Pandiscia claims that Twitch and Amazon began scraping creator content for AI training as early as 2024 without disclosure, only adding an opt-out mechanism after the practice came to light in August. This timing is significant: in 2024, Mike Minton, then Twitch's Chief Monetization Officer, acknowledged the AI training was occurring but framed it as "prototyping, not in any kind of production scale." By August, when Twitch publicly confirmed the practice and added the opt-out setting, Minton (now chief product officer) defended the opt-out-by-default approach, arguing that an opt-in model would result in near-zero participation. Pandiscia argues this admission reveals the company prioritized scale over consent. The lawsuit also exposes a structural loophole: because Twitch's system captures multi-party communications (streams, chats, clips with viewer interaction), obtaining consent from all parties is technically infeasible, yet the company proceeded anyway. This claim challenges whether Twitch's legal framework for AI training can ever satisfy consent requirements when creators and viewers are entangled in a shared digital space.
For example, today's edition would include:
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · takes 30 seconds · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. Q&As are published on this page for other readers too.
Analyst Ming-Chi Kuo says Nvidia has revived the Rubin CPX AI accelerator with a substantially redesigned arch…

A UK study by UK AI Security Institute and Limbic AI surveyed 6,474 British adults

Broadcom's Clayton Donley says companies are doing mission-critical work with AI agents quickly, but without t…
Bank of England governor Andrew Bailey warned that advanced AI poses risks to financial infrastructure in a le…
OpenAI released a new evaluation framework on July 17, 2026, urging companies to measure AI ROI by 'useful out…

As AI agents perform real business tasks, 'Agentic Identity' (giving each AI a unique employee-like ID) and 'D…
