YouTube lets creators control who trains on their videos
YouTube just launched a new opt-in system that gives creators control over which third-party AI companies can train on their videos. The feature defaults to blocking all third-party scraping, offering a whitelist of 18 firms like OpenAI, Anthropic, and Apple. The catch: Google itself will still train its own models on creator content.

If you've spent the last year watching AI video models get freakishly good, there's a strong chance your own YouTube videos helped train them—without your permission and without a dime in return.
On Monday, YouTube finally gave creators a way to draw the line. The video platform rolled out a new dashboard control that lets creators explicitly choose whether third-party companies can train AI models on their videos.
An opt-in list of 18 tech giants
Here's how the system works.
By default, every creator is opted out. Third parties are formally barred from using your video catalog unless you proactively head into your settings and grant permission.
Inside YouTube Studio's Content Manager, administrators can now review a list of 18 major generative AI developers. The initial lineup includes OpenAI, Anthropic, Apple, Meta, Microsoft, Nvidia, ByteDance, Runway, Stability AI, and xAI.
Creators can whitelist specific companies one by one, or simply flip a master switch labeled "All third-party companies" if they don't mind who scrapes their catalog.
YouTube framed this as a foundation for future compensation models. Eventually, the company plans to let approved AI developers directly download source videos from consenting creators.
There is, however, one glaring blind spot: YouTube admitted it cannot retroactively undo the training that third parties have already done on public videos.
The Google-sized loophole
While external tech firms get locked out by default, Google isn't locking itself out.
YouTube confirmed that this toggle applies strictly to third parties. Under existing terms of service, Google will continue training its own internal AI models on YouTube videos as usual.
That detail hit on the exact same day Google DeepMind unveiled Veo 2, its high-end video generation model meant to take on OpenAI's Sora.
Back in September, YouTube first teased these creator controls alongside plans to expand Content ID to detect AI clones of creator voices and likenesses.
Why it matters
We think YouTube's move is clever, opportunistic, and long overdue.
For creators, having an explicit opt-in setting is a meaningful leverage upgrade. Even if no royalty checks are arriving today, turning off permission gives creators legal and moral standing against scrapers.
Yet the hypocrisy is hard to ignore. Google keeps its own private buffet open while shutting the door on competitors under the banner of creator protection.
For rival AI video startups, the era of frictionless web scraping is evaporating fast. If players like Runway or Pika want premium training footage going forward, they will likely have to pay through YouTube's tollbooth.
Ultimately, YouTube isn't just shielding its creators. It is fencing off the world's most valuable video repository to protect Google's own moat.
Sources (1)
- techcrunch.com(opens in a new tab)Originaltechcrunch.com
Topic trackerAI Video Training Data Acquisition
- OpenAI and Google are paying creators $4 a minute for scrap video
- YouTube lets creators control who trains on their videosThis story