Tag: Featured

Thoughts on the No AI FRAUD Act

The acronym stands for No Artificial Intelligence Fake Replicas and Unauthorized Duplication. Introduced as a discussion draft by Rep. Maria Salazar et al., the No AI FRAUD Act would create a novel form of intellectual property in direct response to the use of AI to “clone” a likeness. With parallels to right of publicity (ROP) law, combined with a copyright-like, transferable ownership of rights, the No FRAUD bill is sweeping as currently proposed, citing a range of conduct, from deepfakes to create and distribute nonconsensual intimate material, to cloning an actor or singer’s voice for commercial exploitation.

In short, the law would prohibit replication of anyone’s likeness without permission, and then, the purpose of the unlicensed replication would determine the nature of the harm and available remedies. Although the intent of this bill is well-founded in addressing certain harms to individuals like performing artists, the bill’s current scope, combining permission and intent, and seeking to remedy a broad range of potential harms, raises some difficulties.

Permission vs. Intent

As discussed on this blog, Cyber Civil Rights Initiative (CCRI) leaders, Danielle Citron and Mary Anne Franks, have advocated a permission-based, rather than an intent-based cause of action for the nonconsensual distribution of intimate material, commonly referred to as “revenge porn.”[1] The CCRI has worked hard to demonstrate that merely distributing this material without permission is criminal, regardless of the intent to cause harm, and this makes sense in response to the nature of the conduct. But advancement in AI replication presents a unique challenge to the principle that permission is universally the signal event triggering liability.

No question that the guy who shares intimate material of an ex, a girl at school, a work colleague, etc. should be held accountable solely on the basis that he lacked permission, and this is valid whether the visual material is real (i.e., photographic) or synthetic (i.e., produced with an AI). First Amendment defenses for this type of conduct have reasonably failed when various parties challenged the constitutionality of several of the “revenge porn” laws, now in force in 48 states. The permission principle in harassment-based complaints should not be disturbed by the No FRAUD Act, and Congress should likely avoid any temptation to combine the intent of this bill with current or developing federal prohibitions for “revenge porn.”

But the use of AI to replicate a likeness cannot so broadly be proscribed for all purposes. As the Motion Picture Association notes in its response to the bill, “… any legislation must protect the ability of the MPA’s members and other creators to use digital replicas in contexts that are fully protected by the First Amendment.” Notwithstanding contractual conflicts that may arise in the future among performers and producers, the MPA is right to note that AI cloning for expressive purposes that constitute protected speech should not be swept into the scope of legislation like the No FRAUD Act.

The example I often use with friends and colleagues is the movie or TV series that casts a public figure (let’s call him Donald Trump) in a light he might not appreciate. Expressive portrayals—factual, dramatic, or sardonic—of public figures are paradigmatic forms of protected speech, and this principle should not be altered by vesting new IP rights in persons, premised solely on the use of AI models to achieve the same expressive results historically created with old-school “movie magic.” In other words, Trump should no more be empowered to enjoin the use of his AI likeness to comment upon his role in society than he would have been allowed to stop Saturday Night Live from producing the sketches featuring Alec Baldwin.

Vesting new “likeness IP” rights in all persons is a reasonable response to the potential harms—both financial and reputational—that may be caused to millions of creative professionals and ordinary citizens. But these goals must allow for expressive uses of AI replication, adhering to longstanding contours protecting the speech right and controlling limits like libel and defamation.

In another example, imagine a documentary about the events of January 6th that includes reenactments based on witness testimony describing the actions of the former president during the attack on the Capitol. The documentary producer’s legal responsibility to balance faithful reportage with reasonable expressive license should not be altered solely on the basis that the film may use generated AI likenesses of Trump, Meadows, Hutchinson, Ivanka, et al. rather than actors to produce the same scene.

With a documentary film, one can imagine a legal requirement to inform the viewer that what they are seeing is an AI-generated reenactment (rather than, say, someone’s cellphone recording), but no such requirement should apply to a non-documentary audiovisual work. In either case, misinformation is already thriving in a dangerously blurry space between fact and fiction and a decline in media literacy fostered by the ability of any individual to distribute any fragment of material without context on a public platform. In other words, the documentarian can do her job right, but she cannot stop every potential bad actor from taking a segment of that reenactment and publishing it in a manner that changes its context and feeds a false narrative. (Thank you to all those who celebrated “remix culture” as a rejection of copyright law.)

AI Generated Likeness and the Misinformation Problem

Regarding the documentary example, the preamble of the No FRAUD working draft cites the use of unauthorized likenesses for the purpose of disinforming the public about matters of a factual or newsworthy nature. And while this is indeed a problem that AI tools will be used to exacerbate, it is a challenge that should be addressed separately from the intent and sweep of the No FRAUD proposal. Congress must recognize that the capacity to cause widespread, societal harm through disinformation by means of AI likeness replication is too hazardous and too rampant to remedy on a case-by-case, civil-liability basis. And that’s even if the producer of the fake is operating within the reach of U.S. law rather than, say, China or Russia.

Further, there is a legal tension created by comparing and contrasting the entertainment satirist with the news provocateur who trades in misinformation, as we see in the claims of slander against Tucker Carlson of FOX News in 2020. Arguing that “no reasonable person” would truly believe everything Carlson says, Fox’s attorneys successfully defended the network against any cause of action, and while this may be a reasonable finding based on the facts presented, it is one of many examples in which the lines separating opinion, criticism, satire, and information have been blurred beyond relevance vis-à-vis public perception. Now add the ability to cheaply recreate anyone’s likeness with sophisticated AI, and how far can a “news” organization push the line under the same protections that apply to the satirical filmmaker or The Daily Show?

Of course, my references here to Trump and Carlson allude to a much bigger, underlying problem—namely that Congress is not going to effectively address the use of AI likeness for misinformation unless Members on both sides can agree to mutually define fact and fiction. Not to say that Dems never cling to narratives built on some rather shaky foundations, only that it’s hard to compete with the existential lies of whatever the hell the GOP has become in the thrall of Trumpism. That and no American political figure has ever proven to be so thin-skinned in response to criticism.

For the moment, my own view is that a bill like No FRAUD should be narrowly tailored to vest new “likeness IP” in persons to proscribe compelled speech and commercial exploitation that meets standards akin to unfair competition. Further, because such uses require a court to weigh the intent of likeness replication, this new right should not preempt or alter anti- “revenge porn” legislation, where lack of permission must remain the sole cause of action. While I see the potential of this bill to protect various artists and non-artists with novel rights against novel harms, difficulties like those addressed in this post must help define the contours of those new rights.

[1] “Revenge porn” is a problematic term because it implies intent to harm, which is anathema to the principle that lack of consent is the cause of action.

Image by: meyerandmeyer

“Fair Use” is Not a Great Business Plan

Lately, we’ve seen several headlines and comments from tech giants say that AI ventures simply cannot succeed if they are forced to contend with the copyrights in the billions of works they have scraped for the purpose of machine learning (ML). When these headlines are paired with the rampant assertions that ML is inherently fair use—a subject addressed in last Wednesday’s Senate Judiciary Committee (SJC) hearing on AI and journalism—one has to wonder about the business decisions being made before generative AI exploded last year.

In many posts on this blog, including at least a few written during “Fair Use Week,” I have repeated the caveat that “fair use” is not a magic phrase that makes infringement claims disappear. Usually, that advice is directed at small and independent users of works, suggesting they not listen to Big Tech and its network of academics and activists, who will not be on the hook for the small guy’s copyright infringement. I always assumed the big guys knew better, that they were merely chanting the “fair use” mantra as a rhetorical device in the blogosphere to promote the anti-copyright agenda. But maybe they don’t know better.

If I were an AI investor asking about potential liability, and the founders told me, “Don’t worry, what we’re doing is fair use,” my immediate response would be to ask whether there is sufficient funding for major litigation, to say nothing of predicting the outcome of that litigation. Because simply put, the party who conjures the term “fair use” has effectively assumed that a potential liability for copyright infringement exists. And if that assumption is a bad business decision, then that’s the founders’ problem, not a flaw in copyright law.

No matter what the critics say, or how hard certain academics try to alter its meaning, the courts are clear that fair use is an affirmative defense to a claim of copyright infringement, which means that building a business venture on an assumption of fair use is tantamount to assuming that lawsuits are coming. And if it’s a multi-billion-dollar venture that potentially infringes millions of works owned by major corporations, then the lawsuits are going to be big—perhaps even existential.

Do Not Expect Congress to Change Fair Use in Any Direction

Notably, as reported in Wired, Conde Nast CEO Roger Lynch stated at one point during questioning by the SJC last week, “If Congress could clarify that the use of our content, or other publisher content, for the training and output of AI models is not fair use, then the free market will take care of the rest,” to which Sen. Hawley replied that this seems reasonable. But I wonder about this exchange. While it is encouraging to find the senators more sympathetic with the news organizations than with the AI developers, I doubt (and would not even hope) that Congress is going to amend the law to explicitly state that ML is categorically never fair use.

Fair use comprises a history of judge-made law that was codified into statute as Section 107 of the 1976 revision of the U.S. Copyright Act. But the statute does not draw bright lines stating that X is always fair use and Y is never fair use, and for good reason. Because justice for all parties is best served by a court weighing the specific facts of a specific use of a specific work, or body of works. Hence, an attorney will tell you that fair use is a “fact intensive” consideration.

If Congress were to explicitly declare, for instance, that ML can never be fair use, this would be a significant departure from doctrine, and one that is preemptively unjust to the potential AI developer with a fact pattern that would favor a finding of fair use. As much as I find the major generative AI companies to be some combination of arrogant and/or useless, and as much as I scorn their generalizations to-date about fair use, it would be wrong to endorse legislative revision of the fair use doctrine as a sound response.

In fact, if the court were to find fair use for ML in New York Times v. Open AI (and I doubt it will), and Congress sought to remedy that outcome, it would still not make sense to amend Section 107. If anything, news organizations and other copyright owners would likely seek a new section of the Copyright Act tailored to the nature of the new form of harm, which Big Tech would then blindly oppose with every available resource. For instance, it is possible that the Times would not currently be suing Open AI if the tech industry had not opposed the Journalism Competition and Preservation Act (JCPA), which would have temporarily exempted news organizations from antitrust barriers to collective bargaining for licensing their content.

Regardless, no party should be asking Congress to “clarify fair use” in response to AI. If the AI founders and investors made a bad bet on an ultimate finding of fair use, that’s tough noogies for them. But neither should content creators want Congress to open that particular can of worms and disturb the fair use case law. Of course, where Congress should intervene is to address harms caused by AI where no law currently applies. On that subject, the next post discusses the recently proposed No AI FRAUD Act.

Phot source by areporter.

Recent AI Copyright Lawsuits Are About More than Compensation for Authors

Last week, writer and broadcaster Andrew Keen invited me to his podcast Keen On to talk (of course) about artificial intelligence. When we got to the subject of the New York Times lawsuit against Open AI and Microsoft, I noted that 1) it is arguably the strongest copyright case presented to date against an AI developer; 2) that it would likely result in a substantial licensing deal between the parties; and 3) that it is hard to say what any of this means for journalism going forward. On that same subject, nonfiction authors Nicholas Basbanes and Nicholas Gage filed a class action suit against Open AI and Microsoft on January 5, just over a week after the Times suit was filed.

As discussed in other posts, although generative AI unequivocally poses a threat to authors and authorship, U.S. copyright law is, oddly enough, not quite designed to address the full scope of the social, economic, and cultural challenge of that threat. While this seems counterintuitive, the difficulty lies in the fact that copyright promotes authorship by protecting works against specific means of infringement, and the nail-biting question of the moment is whether “machine learning” (ML) with the use of protected works violates the reproduction right (§106(1)) of the Copyright Act.

Here, the Times case is strong because the news organization presents compelling, side-by-side evidence that its published stories are being output by ChatGPT almost verbatim. This is evidence that not only is reproduction occurring in the AI model, but that the outputs provided to users serve as a substitute for legal access to the Times’s material. The evidence of reproduction establishes a solid claim of infringement, while the evidence of substitution goes against Open AI’s putative fair use defense. In fact, it was the same circuit (the Second) which held that a news service called TVEyes was “slightly transformative” but that it made so much of Fox News’s material available, even in segments, that the substitutional purpose doomed its fair use defense.

Unlike the Times, the nonfiction book authors do not present side-by-side evidence of verbatim copying of their published writings, and this is consistent with some of the other class-action suits. These are the real nail-biter cases, in my view, because the plaintiffs’ cause is just, but their proof of copyright infringement is less demonstrable than the Times (or the Concord v. Anthropic case for that matter). But this focus on both The New York Times and nonfiction authors raises a serious question as to whether AI will exacerbate the already dismal state of information in the information age.

When the early work of this blog started in 2011, one of the issues of concern was the volume of mediocre, careless, or inaccurate reporting and commentary being promulgated under brands normally associated with quality journalism. Here, it must be said that the Gray Lady herself has not always been immune to the digital-age forces of volume and speed that can drive reporters and editors to engage the market on the lowest rungs. But if the stodgy algorithms of social media have animated a new era of yellow journalism, isn’t it reasonable to assume that certain generative AIs will make matters worse? The internet has already fostered more misinformation than a democratic society can safely endure.

If we consider the possible outcomes of the Times lawsuit, one would be that Open AI changes the model to avoid infringing reproduction. While this may satisfy from a copyright perspective, one wonders about the quality and/or purpose of the information being provided by a tool like ChatGPT.  The output of an LLM is the result of probability. The user asks a question (a prompt), and the AI responds that in all likelihood, based on the information fed into an algorithm, this is what you want to know.

It is no wonder the system to date reproduces material verbatim from a major news organization, but if it doesn’t do that, what should it do? Or what can it do that can be called “progress” with regard to news and information? Take a multi-faceted, extremely emotional topic like Israel and Palestine, train an AI on all the solid reporting, all the mediocre editorials, and the cacophony of opinions on social media, and the user of the LLM gets…what? Why would the results be more informative or thoughtful than the veteran journalist doing her best?

Why won’t an AI be worse than “recommendation algorithms?” If YouTube and Facebook foster confirmation bias and shepherd people onto the wild grazing fields of organically grown conspiracies, it seems rational and prudent to assume that an LLM will do the same thing more efficiently. Why have an old-school search engine point you toward a bogus article linking vaccines to autism when you can have a “dialogue” with an ersatz intelligence on the same topic?

Although the nonfiction book authors do not present the kind of evidence of copyright infringement the Times exhibits in its complaint, the facts presented about the authors’ investment of time, expertise, and money makes a point that should be read as more than a mere plea for sympathy. This is not just about job loss for future historians but quite possibly about the loss of history itself.  From the Basbanes et al. complaint:

The archive of primary research materials assembled by Mr. Basbanes in support of his work over a period of forty years, when acquired by Texas A&M University in 2015, filled 365 packing boxes with documents, transcriptions, drafts, field notebooks, photographic negatives, and the like, all acquired by Mr. Basbanes in pursuit of his literary activities, and at his expense and initiative.

It is more than a legal (i.e., fair use) question whether the purpose of a model like ChatGPT is to make new and relevant use of all that work, or whether its purpose is to supplant the historian and the reporter by “feeding off the sere remains of the past,”[1] until it eventually starves. In the former case, licensing and collaborating with authors and journalists seems reasonable, in the latter case, allowing certain generative AIs to die on the vine seems imperative.

[1] From Ralph Waldo Emerson’s speech at Harvard calling for an American literary independence, August 31, 1837.

Photo by: Antonio83