Twitch's product chief on AI training: 'If this was opt-in, nobody would opt in'

Twitch is feeding streamer content to parent company Amazon for generative AI training, and the setting is on unless you turn it off. When streamers asked why the choice runs that direction, Twitch chief product officer Mike Minton answered on a community livestream with the plainest thing anyone in this debate has said out loud.

“

If this was opt-in, nobody would opt in. That's honestly the answer.

TechCrunch reported the exchange on August 12, and Music Ally picked it up the next day, pointing out that artists and songwriters are being handed the same opt-in-or-out decision by their labels and publishers right now.

Where the Twitch opt-out toggle is buried

The toggle lives in channel settings, not the creator dashboard where streamers spend their time. Open channel settings, go to the security and privacy tab, scroll to “training for generative AI,” and switch it off. Twitch documents it in its account settings help page.

Minton could not say whether past streams were already used. Asked directly, he said he did not know what Amazon had done in terms of model training, or what it had and had not used. That answer leaves every streamer with an archive unable to find out where their old content went.

Twitch has framed the existence of an opt-out as listening to its community. Meta offers no equivalent outside the UK, so the bar being cleared here is low.

Why musicians who stream are in scope without being mentioned

The policy is written for content, not for gamers. A production stream, a writing session, a live take of an unreleased song, all of it sits in the same bucket as someone playing Fortnite.

That matters more for a musician than for most streamers. You are broadcasting your voice, your playing, and material you have not released, and voice is the part of a musician’s work with the clearest downstream misuse. Twitch is not asking for those rights separately, and there is no music-specific carve-out in the setting.

The timing is what stands out. Universal’s Lucian Grainge built his AI position on opt-in, the UK dropped its opt-out plan after industry pressure, and SOCAN went the other way with an opt-in attribution model. Music has spent 2 years arguing that consent has to be given rather than assumed. Minton stated the reason platforms resist it, on stream, without being asked twice.

The same gap shows up elsewhere in music tech. Spotify has been unwilling to say what its own models trained on, and session players are still finding out their record contracts never covered AI training. Twitch differs in one way: it told you the toggle exists.