Will LLMs be able to do this legally?

The question has been circling the industry for years, but something about the way it resurfaced on May 8th felt less like speculation and more like a clock audibly ticking. The thread posed it plainly: will LLMs be able to keep doing this legally? The "this" is the foundational act of the entire generative AI industry — ingesting the accumulated written output of human civilisation, without consent or compensation, and distilling it into probability distributions that produce fluent, monetisable text. For years, the companies building these systems operated under a working assumption that training on publicly available data was either clearly legal, or at minimum legal enough to proceed before the courts caught up. That assumption is now under serious pressure. The New York Times lawsuit against OpenAI, the Authors Guild actions, the ongoing disputes with Getty Images, the EU's attempts to retrofit copyright law onto something copyright law was never designed to handle — all of it points in the same direction. The legal scaffolding holding up the training data model was never load-bearing. It was borrowed time dressed up as precedent.

If confirmed as a genuine inflection point, here is what this means. The companies with the largest deployed models — OpenAI, Google, Meta, Anthropic — face potential retroactive liability that no licensing deal can fully extinguish. Data provenance becomes not a compliance checkbox but an existential audit. Smaller players without the legal resources to fight or the capital to settle get squeezed out entirely, which perversely concentrates the market in the hands of the incumbents who caused the problem in the first place. The second-order effect is arguably more consequential: if courts or legislators establish that consent and compensation are required for training data, the cost structure of building frontier models changes overnight. Open-source efforts relying on web-scraped corpora become legally precarious. The entire downstream economy — every startup built on API access to these models — inherits that uncertainty. The internet, which was already retreating behind paywalls and login walls, accelerates that retreat. The data gets scarcer. The models get more expensive. The gap between who can build and who cannot widens into something that looks less like a market and more like an oligopoly with legal moats.

Watch for rulings in the Times v. OpenAI case and any legislative movement in the EU or UK that attempts to define training data as a licensable use — either would force the industry's hand in ways that no amount of corporate positioning can absorb quietly.

NewsHive monitors these sources continuously. All signal titles above link to the original reporting.

Intelligence by NewsHive. Need help navigating what this means for your business? Contact GeekyBee →