Eleven words on an official company livestream did more damage to platform trust than a year of privacy-policy footnotes.

"If it was opt-in, nobody would opt in. That's honestly the answer," said Twitch Chief Product Officer Mike Minton, explaining why creators had been enrolled in AI training without being asked. He added: "So it's going to be on by default."

That is not a leak. That is a product executive describing the design logic on camera.

If you record coaching sessions, and almost everyone now does, this fight is closer to your business than it looks.

What actually happened

On August 12, Twitch introduced a setting letting creators opt out of having their channel content used to train Amazon's generative AI models. The setting was enabled by default, so creators had to turn it off themselves.

When switched on, it covers a channel's livestreams, video-on-demand uploads, clips, chat messages, images and text. Twitch's own FAQ gave the example of audio helping refine speech-to-text models, improving captions "across Amazon."

Twitch's head of community, Mary Kish, told viewers on the same stream: "We don't expect you to be happy or excited about this."

She was right. Streamers flooded Twitch's feedback forum, where the top request under creator features asked the company to make AI features optional, sitting at 15,714 votes when Kotaku reported it. Counts published since put the number of objecting creators above 16,000.

Minton's defense was that this is simply how the internet works. "Almost every content service in the world is on by default," he said, arguing that what Twitch was doing differently was respecting the wish to opt out.

Then it became a lawsuit

On August 20, a class action was filed against Twitch and Amazon in the U.S. District Court for the Northern District of California. Lead plaintiff Warren Pandiscia, a Connecticut streamer, claims the companies harvested creators' streams without permission or compensation, violating creators' implied contracts with Twitch and state unfair competition law.

Notably, he is not claiming copyright infringement. The claims rest on Twitch's agreements with creators and California business law.

And here is the line that should make every coach put down their coffee. Central to the complaint is the argument that once content feeds into an AI system, creators cannot "claw back" what they consider their intellectual property.

The case may end up answering a question nobody in the creator economy has a clean answer to: do a platform's general terms of service give it permission to use your content to build generative AI products?

The part that lands on coaches

Coaching already has a rule about this, and it is newer than most coaches realize.

The 2025 ICF Code of Ethics introduced Standard 2.5 with the rise of artificial intelligence in mind. The current Code took effect on April 1, 2025. The standard requires coaches to meet their ethical and legal obligations "directly and through any technology systems" they use, including AI tools and platforms, meaning confidentiality and informed consent apply to an AI transcription tool exactly as they apply to a face-to-face conversation.

Read that next to Minton's admission. The platform sets the default. The coach carries the obligation.

The guidance stops roughly where practice begins. Writing for the Institute of Coaching, one practitioner noted that Standard 2.5's broad mandate remains the primary public institutional guidance, and that the territory between a one-sentence standard and a detailed framework, which is exactly where most coaches operate, "remains largely unmapped."

The same piece kills a comforting assumption. Coaches who believe "private" or "incognito" AI modes protect sensitive client material are wrong; pasted session notes sit on third-party servers regardless of privacy settings.

Not every vendor behaves the same way, which is the argument for actually reading. Zoom's terms state it does not use customer audio, video, chat, screen sharing or attachments to train Zoom or third-party AI models. That clause exists because users made noise in 2023. Zoom rewrote its terms after criticism to state explicitly that it does not use customer audio, video or chat content for training without consent.

So the homework is unglamorous and overdue: open the settings panel on every tool that touches a client session. Recorder, transcriber, CRM, community platform, webinar host. Find the AI training toggle. Note the date you checked.

Then tell your clients what you found, before they ask you.

Neither Twitch nor Amazon has commented on the lawsuit.