Copyright Can't Save Publishers from AI

Copyright Can't Save Publishers from AI

What a week for generative AI and copyright law. This week was a bad week for Getty Images and publishers opposed to AI training on their content.

The headlines are flying, the legal filings are thick, and the only thing less clear than the future of content ownership is the present. If you're a founder, a creator, or just someone who likes to watch billion-dollar companies fight in court, pour yourself a coffee: the AI copyright endgame is nowhere in sight, but the opening moves are wild.

Getty Retreats, But the War Isn't Over

Getty Images, the grand old gatekeeper of stock photography, has quietly dropped its headline copyright claims against Stability AI in the UK. The core accusation? That Stability used millions of Getty's images - sometimes with watermarks still attached - to train its Stable Diffusion model. Getty's lawyers, after months of posturing, have now abandoned both the "training" and "output similarity" claims. Why? The evidence was weak, the witnesses were weaker, and the UK's jurisdictional quirks didn't help. As one legal observer put it, Getty couldn't show that the AI's outputs "reflect a substantial part of what was created in the images".

But don't mistake this for surrender. Secondary infringement and trademark claims remain, and Getty's $1.7 billion US lawsuit is still live. The message: the legal fog is thick, and the real battle may be jurisdictional, not just factual. Meanwhile, Getty is hedging its bets with its own AI image generator, trained only on its licensed library. If you can't beat 'em, build your own model and charge for it.

Fair Use: Mixed Signals

Across the Atlantic, the US courts are finally starting to draw lines in the sand. Two major rulings dropped this week on whether training large language models (LLMs) on copyrighted works is "fair use." The first, Bartz v. Anthropic, is a landmark: Judge Alsup ruled that training LLMs is "spectacularly transformative" and thus protected as fair use. The logic is simple and, frankly, overdue: just as Google was allowed to copy books to build a search engine, AI companies can study copyrighted works to build new, transformative tools. The court dismissed the idea that LLMs are "infringement machines" just because they can generate new text in the style of the originals.

But then came Kadrey v. Meta, and the waters muddied. Judge Chhabria, while ultimately ruling for Meta due to lack of evidence, opined that training on copyrighted works without a license would be "illegal in most cases." He worried about "market dilution" and AI-generated works competing with originals. The problem? This flips the fair use doctrine on its head, treating "market harm" as the only thing that matters and ignoring the transformative nature of AI training. As the EFF notes, copyright is supposed to encourage new expression - even if that means competition.

The Precedent and the Path Forward

The upshot: the first big AI copyright decision is a win for AI, but the road ahead is anything but clear. The Bartz ruling gives AI companies a strong fair use precedent, but the specter of appeals, class actions, and conflicting opinions looms large. Congress is now on the clock to clarify what's allowed and what isn't, because the courts are sending mixed signals.

Meanwhile, the industry isn't waiting. The Big Five tech companies are adapting, hedging, and investing. Getty is building its own models. Apple is betting on private data and device integration. The only certainty is that the legal and business landscape will keep shifting, and the only thing more expensive than training an LLM is litigating one.

Second Blow: The AI Browser Wars and the Coming Crisis for Publishers

If the legal fog wasn't enough, this week delivered a second seismic shock - one that should have every publisher dependent on search engine traffic reaching for the antacids. The AI browser wars have begun, and the implications for the open web - and the revenue models that depend on it - are profound according to MG Siegler

The launch of Dia, an AI-first browser from The Browser Company, marks a turning point. Unlike legacy browsers, which are hurriedly bolting on AI features, Dia is built from the ground up to make AI the default layer of the browsing experience. With integrated chatbots, instant page summaries, and the ability to query across all open tabs, users are nudged to interact with content through AI intermediaries rather than directly with publisher sites.

Why does this matter? Because the traditional web ecosystem is built on a simple value exchange: publishers create content, search engines index it, and browsers deliver users - who then see ads or hit paywalls. If AI browsers become the primary interface, users may never even visit publisher sites. Instead, they'll get AI-generated summaries, answers, and context - sidestepping the original source entirely.

This isn't just a theoretical risk. The article notes that, for many, the AI overlay is already becoming the default way to consume news, research, and information. Chrome's integration of Gemini is clunky for now, but Google's ambitions are clear. Meanwhile, upstarts like Dia and Perplexity are racing to own the AI browser experience, giving them unprecedented control over what users see, how they see it, and - crucially - what they don't.

For publishers, this is a direct threat to both traffic and revenue. The "click-through" model that underpins digital advertising is at risk of being bypassed. If users no longer need to visit a site to get the information they want, ad impressions and subscriptions will plummet. The browser, once a neutral conduit, is becoming an active filter and gatekeeper - one that may not have publishers' interests at heart.

The bottom line: the AI browser wars are not just a product story - they are an existential challenge to the business model of the open web. Publishers who rely on search-driven traffic must now grapple with a future where the browser itself intermediates, summarizes, and potentially monetizes their content - without ever sending a user their way.

Philosophical Coda: The Internet of Tolls - A New Commons or a Fragmented Future?

As the dust settles on a week of legal wrangling and technological upheaval, a deeper question emerges: What kind of digital society are we building as AI, browsers, and copyright law collide? The answer, it seems, may be found at the tollbooth.

Om Malik and Fred Vogelstein's reflection on the coming "Internet of Tolls" captures the sense that the open, frictionless web - the commons that fueled decades of innovation and free expression - is fragmenting. The new AI-powered landscape is increasingly shaped by paywalls, subscriptions, licensing deals, and, yes, literal and metaphorical tollbooths. Publishers, squeezed by search disruption and AI summarization, are racing to erect barriers around their content. Platforms, in turn, are building their own walled gardens, seeking to capture value before it leaks out into the ether of generative models.

But as Crazy Stupid Tech argues, these tollbooths are not just about money - they are about power, control, and the ability to set the rules of engagement for the next era of the internet.

The AI revolution, for all its promise, risks turning the web into a patchwork of private roads, each with its own gatekeeper, fee structure, and terms of service. The dream of a universal, open commons is giving way to a reality of fragmented access and algorithmic mediation.

Is this inevitable? Perhaps. As value concentrates in the hands of those who control the infrastructure - whether it's the browser, the model, or the content feed - the incentives to charge, restrict, and gate will only grow. Yet, as both articles remind us, the real challenge is not technological but philosophical: Will we use AI to reinforce the walls, or to build new bridges? Will the internet become a maze of tolls, or can we imagine new models of sharing, compensation, and creativity that keep the commons alive? Cloudflare seems to be thinking about that (see Post of the Week).

The week's events - AI browser wars, copyright battles, and the rise of digital tollbooths - are not just business stories. They are signals of a deeper transformation, one that will define who gets to participate, profit, and create in the digital age. The choices we make now, about law, technology, and values, will determine whether the future of knowledge is open or closed, abundant or fenced off.

That was the week. The tollbooths are rising. The road ahead is ours to choose