The American AI and Copyright Accord
What matters more to America’s future: leading the AI revolution that will shape our future or protecting the constitutional right of copyright that has built our creative and economic strength? The answer is both.
In courtrooms across the U.S., AI companies and creators are clashing over how copyrighted works are used to train AI models.
Creators see their life’s work used without permission or compensation, despite the fact that copyright has been enshrined in Article I of the Constitution since our country’s founding to protect and encourage innovation and ownership.
AI developers race to build world-changing technology, stating that any obstacles to training could make America fall behind global rivals.
The legal landscape is divided. The outcomes of some of these lawsuits are starting to emerge, but the final word may likely be decided by the Supreme Court.
At the center is a flawed argument that an AI system training on a work is the same as a human reading and learning from it. But this analogy misses the mark. Sure, people read books and are influenced by them. But no human can comprehend millions of titles and instantly generate hundreds of millions of outputs drawn from them in seconds…all while monetizing those outputs at scale.
This isn’t fair use as we've known it. It’s something entirely new. And it demands new rules.
Author Isaac Asimov understood this tension long before our current era. In “I, Robot,” he introduced the “Three Laws of Robotics” to guide a world in which humans and robots coexisted.
But Asimov’s fictional story imagined robots as physical beings, separate from us. We need a new guiding set of principles for our reality, where humans and machines co-exist and create together.
We propose building upon his work with the Fourth Law: When AI learns from human creative work, it must do so symbiotically—benefiting both humans and AI. This means that humans should have the right to consent to their content being used, maintain some control over how it’s used, and have the right to be compensated for it. It’s time we establish these rules between human and robot interactions, just as Asimov suggested decades ago.
Another argument often made is that America must relax copyright protections to stay ahead of China in the AI race. But gutting our own laws and industry leadership isn’t strategy, it’s surrender. We don’t win by abandoning the creative rights that have defined our economy and culture since its inception. Developing a system that protects human creativity and pushes AI innovation gives us the competitive advantage.
Introducing the American AI and Copyright Accord
Following the Fourth Law, we can protect creators and supercharge American AI innovation if we modernize the rules and build a licensing system that works at machine speed.
We've consulted with bestselling authors, AI developers, artists, publishers, researchers, and technology executives to develop a clear, fair, and scalable blueprint that is built on three principles:
- Accelerate American AI innovation
- Ensure creators are fairly paid and the creative economy in the U.S grows
- Keep it simple and scalable
Done properly, artists and other rights holders get paid, AI companies get the highest quality data they need to build the best models, and America gains the moral and legal authority to ban AI models built upon stolen IP, such as China’s DeepSeek.
The promise of a true win-win-win is real: faster innovation, stronger creative industries, and enduring U.S. leadership. But promises require action. If we want a system that works for everyone, now’s the time to build it intentionally, and together.
How we'll do it:
5 steps to make copyright work for AI
STEP 1
Treat the open web and copyrighted commercial works differently.
Most copyrighted web content is publicly available but lacks clear guidance for AI crawlers on whether it’s available for use without a license. By evolving the robots.txt protocol into an AI robots.txt standard, websites can specify what’s available and what’s not with a few lines of code. This could free up the majority of the web to be available for training without a license, while ensuring that commercially copyrighted works (like books, journalism, music, art, and more) are clearly marked as opt-in only and accessible through licenses.
Step 2
Unlock scalable licensing through AI Rights marketplaces.
AI companies need a handful of clean batch licenses and instant access to content, not thousands of fragmented deals. Rights holders need a simple way to opt in and receive royalties. A growing ecosystem of AI Rights Marketplaces can meet both needs. These new platforms will aggregate creator permissions, issue standard licenses, offer efficient content access to developers, and distribute royalties through a clear system.
STEP 3
Licensing for AI training should be driven by market dynamics, not static rates.
Developers and rights holders can negotiate pricing based on content type, scale of use, and system architecture.
Alternatively, collective licensing models, similar to those used in music, can offer predictable access based on factors like the volume of copyrighted content used, the developer’s topline revenue, and organization type (e.g., commercial, nonprofit, educational).
Both paths ensure that rights-cleared content remains accessible at scale, while ensuring fair, transparent compensation for creators.
STEP 4
Clarify AI use: training, reference, and transformative.
We need to distinguish between a model learning from a work, referencing a work, and generating new content based on that work.
- Training can be addressed through the revenue-share model
- Reference uses, like RAG, should include proper attribution and reflect actual usage
- Transformative uses that change the original works into something new require bespoke licenses
Making these distinctions clear helps everyone understand the rules.
STEP 5
Create a good-faith safe harbor.
Going forward, inadvertent use of unlicensed content will be protected, provided the company demonstrates good-faith licensing, filtering, and retroactive compensation procedures. This will require legislation and could resemble the Digital Millennium Copyright Act’s notice-and-takedown system, but is updated for the scale and speed of modern AI.
Why it works
AI developers worry about managing licensing complexity and speed. A revenue share with an intermediary who handles the licensing and payout details solves this.
Rights holders worry about erosion of value. A guaranteed revenue stream reduces that fear.
Each side gives a little and gets certainty in return.
Why now?
A rights-cleared AI stack is a strategic advantage for the United States.
A clear, U.S.-led licensing framework becomes an exportable model, allowing American AI firms to out-compete foreign rivals on legal clarity and ethical AI deployment.
Foreign AI developers are building models on IP taken from the U.S. They offer no reciprocity. If we fail to enforce our own IP standards, we legitimize the global exfiltration of American creative capital.
But strong copyright protection for AI flips that script. It becomes both a legal shield and a diplomatic tool. It gives the U.S. the moral and legal authority to ban or restrict AI models, like DeepSeek, trained on stolen content.
The result is an AI economy that advances U.S. technology, values, and security.
Our roadmap
The roadmap is simple: develop clear web standards and competitive rights marketplaces. The industry competes under it, and creators get paid. Adopt it now and the United States scales AI faster and more fairly than any other country.
Let’s act to keep America at the frontier of both technology and creativity.
Join the movement to protect human creativity and advance AI.
Add your name in support of the American AI and Copyright Accord.
Add your name