Why Paying Copyright Holders for AI Training is Essential

Using someone’s copyrighted material without permission or compensation isn’t innovation. It’s exploitation.

As discussions around AI development accelerate, it’s vital to understand how these systems are built and why the creative economy that fuels our media, culture, and technology deserves protection and support.

AI Doesn’t Appear Out of Thin Air

Large language models and image generators aren’t born with knowledge. They’re trained on enormous datasets that often include books, news articles, film scripts, artworks, blog posts, and more - often without the consent, awareness, or compensation of the original creator. The outputs may be impressive, but the foundation of these models is built on the work of real people.

The reality is that scraping the internet for AI training has become an unstoppable force, and the sheer scale of publicly available content, combined with a global race to build ever-larger models, ensures that this practice will continue, regardless of its legality. TollBit data shows AI scraping traffic volume nearly doubled in Q1, rising by 87% during the quarter.

That raises the question no longer of whether scraping can be stopped, but how we can impose guardrails, enforce accountability, and ensure creators are fairly compensated in a system that is already racing ahead.

You Can’t Build the Future by Ignoring the Past

Paying creators for the use of their work is not just about fairness - it’s about sustainability. Copyright law exists to recognize and protect creative labor. Writers, artists, journalists, musicians, and filmmakers rely on this system to earn a living.

The AI industry now faces its own responsibilities, following the music industry's transition from Napster to Spotify and the publishing industry's shift to eBooks and digital subscriptions. The Streaming Media industry, regardless of content, pays licensing fees to its content providers. The payment requirements for AI should be no exception. Publishers are proactively trying to address the issue. Tollbit has seen a 4x increase in publishers attempting to block bots.

When AI Trains on AI, Everyone Loses

An often overlooked risk is what happens when AI models begin training on outputs generated by other AI systems. This feedback loop can degrade the quality, originality, and diversity of future content. Over time, we risk producing derivative and homogenized outputs that lack the depth, nuance, and creativity that only human experience can provide.

Preserving original human-created content and ensuring it remains part of the AI training ecosystem is essential for the long-term health of both the technology and the broader information economy.

Powering a Future With Consent and Creativity

There’s no reason AI development and creator rights need to be at odds. Many creators are open to licensing their work - provided they’re offered clear terms and fair compensation. Tools like TollBit and Spawning.ai show that consent-based, licensed training models are not only feasible but already in motion.

Obviously, this isn’t about slowing down innovation - it’s about strengthening it by ensuring the people who make the internet meaningful are included, respected, and rewarded.

Paying copyright holders is foundational to building a future where AI is both powerful and principled.