“If it was opt-in, nobody would opt in”: Twitch auto-enrolls all streamers in having their content scraped for Amazon LLM training

Despite significant backlash and clear communication from streamers, Twitch has gone ahead with plans to scrape their content to train Amazon‘s LLMs.

As of Aug. 12, Twitch auto-enrolls all channels in consenting to have their content scraped for training. People who do not actually consent have to go into their settings and use a toggle to opt out.

Why did Twitch decide to make scraping opt-out instead of opt-in? CPO Mike Minton gave a “jaw-dropping” explanation on Twitch’s latest Patch Notes stream: “If it was opt-in, nobody would opt in.”

“That’s honestly the answer,” he added, as his stream cohost, Twitch’s Head of Community Mary Kish, winced.

Minton also argued that companies like OpenAI and Anthropic already scraped the entire internet to train their models, so Amazon is just following that precedent and not really doing any additional harm.

“If models were trained on your VOD, it may or may not even have been Amazon,” he said. “It was likely any one of the […] dozens of companies that are creating models that are using any publicly available content they can find.”

He gave vague answers to chatters’ concerns over whether Amazon might have previously scraped streamers’ content without telling anyone. “Twitch has not been training models,” he said. “I can’t speak to what Amazon is training or not training as it relates to any specific usage of content.”

In the stream’s chat, Kish attempted damage control, saying, “The opt out numbers will help us continue to reiterate that our community is not receptive of AI in any form. I’m sorry but I am reading your comments and do hear you.”

The one slight glimmer for streamers here is that, unlike YouTube or TikTok, Twitch does not keep permanent libraries of streamers’ content. Broadcast VODs are auto-deleted at 60 days maximum–meaning if a streamer opts out of scraping tomorrow, the most Amazon would get is their last 60 days of content, versus years of videos.

Backlash against this decision extends beyond the Patch Notes chat. The current top post in the Feature Requests section of Twitch’s User Voice forum asks the platform to make scraping opt-in. As Aftermath first noted, that post has more than 13,000 upvotes and thousands of supportive comments, compared to ~250 votes for other high-traffic suggestions.

This is, once again, a consequence of major digital content platforms all being owned by tech giants. We’ve extensively covered the interplay between companies like Google, Meta, ByteDance, and Snap all rushing to make money on the AI bubble, while YouTube, Facebook/Instagram, TikTok, and Snapchat end up overrun with generated slop.

In many cases, these companies frame their AI push (including potential content scraping) as a boon to creators, with marketing language that ignores any protest and suggests AI advancements are inevitably ‘supercharging’ human creativity in content.

Amazon and Twitch are the first to frame things in such bald terms: “If it was opt-in, nobody would opt in.”