Twitch has introduced a new privacy setting that lets streamers opt out of having their content used to train generative artificial intelligence models developed by Amazon, Twitch’s parent company.
The update has reassured some creators, but Twitch has not clarified when it began using users’ posts, livestreams, and videos to train Amazon’s AI systems. The disclosure has renewed concerns about how major technology companies collect and use user data for artificial intelligence training.
Enabling the Twitch AI opt-out is straightforward. On the Twitch website or mobile app, select your account avatar, open Settings, and navigate to the Security and Privacy section. Under Generative AI Training, you can turn off the option that allows Twitch to use your content for AI training.
However, Twitch’s account settings guidance warns that disabling the feature does not prevent Twitch or Amazon from using channel content for other purposes described in Twitch’s Privacy Notice. Those uses include AI-powered tools intended to help streamers grow and monetize their channels, such as real-time sponsorship assistance, audience recommendations, and community-safety features like AutoMod.
The new setting is designed to give creators more control over how their livestreams, videos, and other content are used. Its rollout has also prompted questions about whether Twitch and Amazon have already been using creator content to train AI models.
More than 16,000 creators participating in Twitch’s creator forums have objected to their content being used by default to train Amazon’s AI systems. Many said they became aware of the practice only after the platform updated its account settings.
The backlash followed a livestream in which Mary Kish, Twitch’s head of community, discussed the changes. Twitch executives acknowledged that the announcement could trigger a negative response. Mike Minton, Twitch’s head of product, said the AI-training option was enabled by default because, otherwise, “no one will participate” in the process.
Minton also argued that Twitch is not unique in dealing with this issue. He suggested that other companies developing AI systems may be collecting content from Twitch and other online services to train their models. “We don’t know for sure, but I think it’s very reasonable to think that almost all content that’s published is being used to train models in some way, with or without permission,” he said. “We also need to recognize that there’s a lot here that’s beyond our direct control.”
Those comments raised additional questions. When did Twitch begin using creator content to train AI models? Is Amazon the only company accessing the data, or are business partners involved? And how will Twitch and Amazon protect the copyrights of creators whose work is included in training datasets?
Twitch’s Terms of Service state that, beginning in March 2024, creators grant Twitch and its sublicensees the right to use, reproduce, modify, adapt, distribute, and create derivative works from their content. However, the terms do not explicitly state that Twitch content may be used to train generative AI models.
The AI training data challenge
The Twitch controversy highlights one of the biggest challenges facing AI developers: the growing shortage of high-quality data for training increasingly advanced models.
Companies such as OpenAI have signed agreements with publishers, including WIRED’s parent company, Condé Nast, to license content for AI training. As AI technology advances and demand for data grows, traditional sources are becoming more limited. That scarcity has encouraged companies to pursue alternative datasets while reigniting debates about consent, copyright, transparency, and the ethics of AI training.
Source: www.wired.com