AI vs Copyright: The UK government has picked its side. Here’s why that’s a problem.

A new consultation (open until Feb 25, 2025) proposes exempting AI developers from key copyright protections. Instead of asking permission, developers could use creative works by default—forcing artists, writers, and musicians to opt out if they don’t want their work used.

The government’s logic? The UK is falling behind in AI development compared to the US and China, and loosening copyright rules might help attract investment, talent, and growth.

But here’s the problem: this approach prioritises speed over fairness—and it risks undermining the very creators AI models depend on.

What’s Changing?

Data scraping isn’t new. Social platforms have been mined for years—most of it public content people don’t view as valuable (even if it is). But generative AI is different.

It needs high-quality creative input—art, videos, books, and music—to train better models. And scraping this kind of data is already causing friction. Case in point: in the US, OpenAI and Google are facing lawsuits from YouTubers and publishers for using their content to train AI models without permission.

Now, the UK is proposing to make this easier for developers—by stripping creatives of the right to opt-in and forcing them to opt out instead.

The Real Issue? Compensation—Not Just Copyright.

AI models are designed for profit. But the value chain starts with creatives—whose work trains these systems. Without their talent, these models wouldn’t exist. Yet the proposal ignores this value exchange entirely.

This isn’t just a copyright debate. It’s about ensuring creators get paid for enabling AI businesses to thrive.

But Isn’t AI Output Original?

This is where confusion creeps in—so let’s clear it up. Generative AI doesn’t simply copy and paste. It learns patterns—styles, structures, lighting, and techniques—just like a human artist does before they create something new.

But here’s the catch: the AI still relies on human-created work to learn those patterns.

The result? AI outputs might be new, but they’re also built on existing talent. So originality doesn’t excuse the need to pay back the source.

Existing Solutions—if protecting copyright is the main concern

Anthropic recently settled a copyright lawsuit with music publishers by agreeing to build guardrails into its AI systems to prevent outputs that reproduce copyrighted song lyrics. These measures include filtering content, embedding metadata, and applying protections across both current and future AI models.

This proves AI developers can take responsibility for preventing infringement—without shifting the burden onto creators. So why not scale solutions like this instead of dismantling copyright protections?

A Better Way Forward 🚀

I inherently believe that the problem is not just one of copyright but also fair compensation, given that the trained AI models can directly compete with the same creatives whose work trained the model.

We need solutions that balance AI growth and fairness. Instead of an opt-out free-for-all, the UK could:

1. Enable Opt-In Licensing Platforms – Models like Calliope Networks (in the US) already help creators license content for AI training. These groups could handle licensing and compensation on behalf of creators to avoid operational bottlenecks.

2. Push for Equity-Based Deals – AI developers could grant rights holders equity shares—allowing creators to benefit as AI companies grow.

3. Negotiate Collective Agreements – Groups could handle licensing and compensation on behalf of creators to avoid operational bottlenecks. Existing organisations like SACEM (Société des Auteurs, Compositeurs et Editeurs de Musique) could step in for this.

These approaches protect creators while keeping AI innovation on track.

The Bottom Line:

AI growth shouldn’t come at the expense of fairness. The UK government’s proposal doesn’t strike a balance—it hands developers a blank check to use creative work without consent or payment.

We need better solutions—because progress and fairness don’t have to be at odds.