"If This Was Opt-In, Nobody Would Opt In." Twitch's Product Chief Just Said the Quiet Part.
· CX Pulse
Twitch now trains Amazon's AI on creator content by default. Its product chief explained why it isn't opt-in, and the answer was honest.
"If This Was Opt-In, Nobody Would Opt In." Twitch's Product Chief Just Said the Quiet Part.
Twitch added a setting yesterday that lets creators stop Amazon training generative AI models on their channel content. Streams, videos, clips, chat.
Everybody is switched on by default. To turn it off you go into channel settings, open the Security and Privacy tab, and scroll until you find "training for generative AI."
Asked why it wasn't opt-in, Twitch's Chief Product Officer Mike Minton answered plainly: "If this was opt-in, nobody would opt in. That's honestly the answer."
Give him credit for not dressing it up. Most companies would have said something about improving the creator experience.
The second quote is the one that should bother you
Asked whether Amazon had already trained on Twitch content before the toggle existed, Minton said: "I don't actually know the answer to that question because I don't know what Amazon has done in terms of model training and what they've used and not used."
The person responsible for the product does not know whether the thing the setting protects against has already happened.
That isn't him being evasive. It reads like an honest answer about a very large company where the platform and the model team are different worlds. Which is precisely the problem, and it's the shape of this everywhere.
What this means for you
You probably don't stream. This still lands on you, and it takes one sentence to explain: everything your business publishes on a platform you don't own sits under terms that can change to opt-out-by-default, and the announcement is a settings page you have never visited.
Think about where your business actually puts words and pictures. Your Facebook page. LinkedIn. Google Business Profile. Your reviews. Your marketplace listings. Support conversations inside a helpdesk tool. Recorded sales calls. Every one of those is content sitting on somebody else's platform, under terms you agreed to once and have not read since.
The default is the product decision. Minton said out loud what the whole industry knows: almost nobody changes a default. When a company chooses opt-out, they have chosen the outcome. Treat the default as the policy, because it is.
"We'll let you opt out" is not the same as asking. Consent that requires you to go looking for a scroll-down toggle in a privacy tab is consent in a technical sense only. That's a fair thing to notice, and also a fair thing to remember when you're the one writing terms for your own customers.
Retroactivity is where the real question lives. The setting governs future training. Whatever has already been ingested has already been ingested, and nobody at the company could tell a reporter otherwise. Any opt-out you exercise anywhere is a line drawn from today, never backwards.
What this isn't
This isn't a reason to pull your business off platforms. That advice would be useless and you'd be right to ignore it. Reach lives where the people are.
It also isn't unique to Amazon. Similar defaults have been appearing across a lot of tools this year, usually announced the same quiet way.
The useful reaction is narrower. Know which of your content is training something, decide whether you mind, and stop being surprised by it.
The five minute version
Pick the three platforms your business publishes on most. Open the privacy or data settings on each and look for anything with "AI," "training," or "model improvement" in the label.
You'll find toggles you have never seen. Some you'll want off. Some you genuinely won't care about, and that's a legitimate answer too.
You don't need to switch everything off. You do need to stop finding out from a news story what your own accounts have been set to for months.
Source: TechCrunch, "Amazon will train on Twitch streamers' content by default, unless they opt out," 12 August 2026.