Photo Credit: Steve Johnson
The Trump administration has reiterated its belief that training AIs on protected IP constitutes fair use – while urging Congress to let the courts “resolve this issue.” Additionally, the White House is calling on lawmakers to establish a system through which rightsholders can seek licensing compensation.
We touched on these and other “legislative recommendations” when they released last week. And as we noted, the “National Policy Framework for Artificial Intelligence,” in keeping with the title, describes a comprehensive collection of federal (not state) AI regulations.
Just in passing, this includes a variety of safeguards for minors (and a guarantee that states can still enforce “their own generally applicable laws protecting children”), steps to curb AI-powered impersonation scams, programs to encourage AI adoption among small businesses, First Amendment protections, and much more.
Unsurprisingly, then, the proposal also addresses the all-important training question – beginning with the aforementioned deferral to the courts in terms of determining what constitutes fair use.
“Although the Administration believes that training of AI models on copyrighted material does not violate copyright laws,” the relevant section reads, “it acknowledges arguments to the contrary exist and therefore supports allowing the Courts to resolve this issue. Similarly, Congress should not take any actions that would impact the judiciary’s resolution of whether training on copyrighted material constitutes fair use.”
As many will recognize, we could explore this multifaceted subject in detail if so inclined; the courts definitely haven’t reached a consensus about the legality of training LLMs on protected materials without prior authorization.
“In short,” Judge William Alsup wrote last year, “the purpose and character of using copyrighted works to train LLMs to generate new text was quintessentially transformative. Like any reader aspiring to be a writer, Anthropic’s LLMs trained upon works not to race ahead and replicate or supplant them — but to turn a hard corner and create something different.”
“There is certainly no rule that when your use of a protected work is ‘transformative,’ this automatically inoculates you from a claim of copyright infringement,” Judge Vince Chhabria penned in an order dismissing an infringement action against Meta. “And here, copying the protected works, however transformative, involves the creation of a product with the ability to severely harm the market for the works being copied, and thus severely undermine the incentive for human beings to create.
“And, as should now be clear,” Judge Chhabria continued, “this ruling does not stand for the proposition that Meta’s use of copyrighted materials to train its language models is lawful. It stands only for the proposition that these plaintiffs made the wrong arguments and failed to develop a record in support of the right one.”
Put differently, training-related infringement litigation’s been a mixed bag for rightsholders.
On the other hand, as demonstrated by Hollywood’s successful clampdown on ByteDance’s Seedance 2.0 video generator, the same rightsholders do have the sway to influence AI product releases in far-flung foreign markets with less robust IP protections – compensation for the underlying training aside.
At the intersection of all those points, can congressional guidance enable rightsholder compensation while simultaneously preventing expensive litigation and uncertainty? The White House believes so – though this guidance wouldn’t directly address whether licensing is required under the law.
“Congress should consider enabling licensing frameworks or collective rights systems for rights holders to collectively negotiate compensation from AI providers, without incurring antitrust liability,” the appropriate text explains. “Any such legislation, however, should not address when or whether such licensing is required.”
In any event, the “licensing frameworks or collective rights systems” would exist alongside and possibly adjust based upon legal decisions, according to the Trump administration’s suggestions.
“Congress should continue to carefully monitor the development of copyright precedents and enforcement in the courts and evaluate whether, due to novel AI considerations, additional action beyond that proposed here is needed to fill potential gaps or provide additional protections for content creators,” the document continues.
Time will tell whether congressmembers rise to the occasion and produce a cohesive AI policy – an admittedly tall task for a legislative branch that has yet to pass the clear-cut No Fakes Act.
(The White House suggestions back “a federal framework protecting individuals from the unauthorized distribution or commercial use of AI-generated digital replicas of their voice, likeness, or other identifiable attributes.”)
Closing with an overview of the AI litigation landscape: Suno recently won a sub-battle to pause discovery in a copyright class action levied by musicians, while counsel in authors’ Anthropic suit formally requested legal fees equaling 12.5% of the $1.5 billion settlement fund (plus a few additional payments).
Technically, said settlement fund will end claims stemming from Anthropic’s alleged book piracy – not its actual training practices, which, as mentioned, Judge Alsup found to be fair use. The judge also deemed breaking down and scanning physical books for training to be fair use; it seems safe to assume that Anthropic, now facing a massive complaint from BMG, has put its alleged file-sharing days in the rearview.
