The Trump administration has been very active in AI policy since he assumed office. Nearly a dozen Executive Orders have emanated from the White House, and on July 23rd the President gave a speech introducing the administration’s 28-page “AI Action Plan.”
There’s a lot of good in the plan, and this post is not intended to be a detailed critique of the plan. A federal pre-emption of state AI regulation is an important feature, although the President can’t just declare it—that requires federal legislation. A provision banning state AI regulation was in the penultimate version of the “One Big Beautiful Bill Act” (OBBBA) but was stricken at the last minute at the request of Tennessee Sen. Marsha Blackburn.
Copyright protection has been a hallmark of Sen. Blackburn’s public service, and for good reasons. Not only does represent the songwriting capital of the world, Nashville, but copyright protection is a critically important area of policy. Trump is right, and Sen. Blackburn is wrong, on federal preemption of state AI regulation.
We at IPI take a backseat to no one in the defense of copyright, a topic we’ve covered for almost two decades. IPI is an accredited NGO with the World Intellectual Property Organization (WIPO) in Geneva, Switzerland, and for several years IPI sponsored the leading World IP Day (April 26) policy event in Washington DC.
So we’re big on copyright.
All of copyright is a balancing act between the rights of creators and ease of public access (notice I didn’t say “the rights of consumers,” because no one has a right to someone else’s property). But there can’t be any discussion of balance until the fundamental right of creators to own and control their creations is acknowledged, and certainly not disregarded.
That’s why President Trump’s comments on copyright and AI are so troubling. No one expects Trump to be an expert on intellectual property law, but the sentiments he expressed would be very dangerous to the U.S. economy.
President Trump said that we can’t expect copyright holders to be compensated for their work by AI companies because “China’s not doing it.” Well, there’s a lot of individual rights that China doesn’t respect, but we aren’t tossing our individual rights because of competition with China.
Trump also seemingly belittled the idea that every time someone reads a book, they have to pay something for it. But, actually, usually we do. You may have bought the book, or someone else bought the book and gave it to you. Perhaps you borrowed it from a library, in which case the library paid for the book under a licensing arrangement with the publisher.
If you’ve viewing a book or article online, chances are you are either paying for it with a subscription, or paying for it by subjecting your eyeballs to advertisements. We do, in fact, almost always pay something in order to read a book or an article.
This post is not intended to be a detailed prescription for balancing the rights of copyright owners with the AI innovators. That’s going to take a lot of work and probably a number of court cases.
But here’s is what is obvious:
- It cannot be that, because China doesn’t respect property rights, neither should we if we want to compete.
- It cannot be that the rights of one set of creators get tossed in favor of the rights of another set of creators.
- It cannot be that we dispose of a proven system of rights because the demands of an emerging technology.
There has always been a tension between copyright and technology, going all the way back to the printing press, but especially since the emergence of the internet. We’ve had to figure out the balance between rights and search engine indexing, rights and streaming, rights and music sharing, and especially piracy. It should be no surprise that there is a tension between those training large language models (LLMs) and copyright holders.
It's a balance that we will be able to solve.
And inputs into LLM training is not the only tension between AI and copyright. Buckle up.