Disney's Billion-Dollar OpenAI Deal Raises the Question of Who Gets Locked Out
Disney’s investment in OpenAI, reported at around a billion dollars, has been discussed mostly as a content deal. Legal scholars looking at it have raised a different concern, and it deserves more attention than the headline number.
The worry is enclosure. When the largest content owners sign exclusive or near-exclusive arrangements with the largest AI companies, the effect isn’t just a transfer of money. It removes material from the pool available to everyone else and raises the entry price for anyone trying to build a competing system.
Kristelia García’s phrase for it is a digital enclosure loop, and the loop is the important part. Big content pairs with big AI. The pairing makes both harder to compete with. Harder to compete with means the next pairing happens on even more concentrated terms.
How it works in practice
An AI company without a licensing budget has two options. Scrape and accept legal exposure, or build with less material and ship a worse product.
Neither is survivable at scale, which means the ability to pay for content becomes a gate on entering the market at all. That gate didn’t exist three years ago, when everyone trained on scraped web text and the legal risk was theoretical.
From the content side the mirror applies. A studio, publisher or catalogue owner that can command a nine-figure deal gets one. Everyone else gets the standard programme, the revenue-share pool, or nothing. The rights market stratifies along exactly the same line as the AI market, and the two reinforce each other.
Why this isn’t the usual complaint about big companies
There’s a standard objection to concentration arguments, which is that scale often delivers better products and lower prices, and consumers don’t care about market structure.
That objection has force in most industries. It has less here, because the product is an information intermediary. If a small number of systems become the primary way people get answers, and those systems are built on material licensed from a small number of owners, the range of what gets surfaced narrows in a way that has nothing to do with price or quality.
It isn’t censorship and nobody intends it. It’s just that a licensed corpus reflects who could afford to be in the room, and the people who couldn’t afford it don’t show up in the answers.
The countervailing case
Worth stating the other side properly. Licensing is also the mechanism by which creators finally get paid for material that was previously taken for free. Every deal signed makes the next unlicensed scrape harder to defend and establishes that content has a price.
An industry where AI companies pay rights holders is better than one where they don’t, and the people arguing loudest against consolidation are frequently the same people who spent two years demanding compensation. You can’t have a licensing market without licensing deals, and licensing deals get made by whoever has leverage.
The concern isn’t that deals exist. It’s that the terms are private, the scale is enormous, and the structure emerging looks a lot like the one the open web was supposed to have replaced.
What would actually change it
Collective licensing bodies with the authority to negotiate on behalf of small rights holders. Standardised terms that a mid-size publisher can accept without a legal department. Transparency requirements on what a licensed corpus contains, so the shape of the enclosure is at least visible.
None of that is close. Meanwhile the deals keep getting signed, and each one is individually reasonable.
That’s how enclosures always happen.