Although that statement may provide comfort to those who have kept their WhatsApp numbers off the Internet, it doesn’t resolve the issue of WhatsApp’s AI helper potentially randomly generating a real person’s private number that may be a few digits off from the business contact information WhatsApp users are seeking.
Expert pushes for chatbot design tweaks
AI companies have recently been grappling with the problem of chatbots being programmed to tell users what they want to hear, instead of providing accurate information. Not only are users sick of “overly flattering” chatbot responses—potentially reinforcing users’ poor decisions—but the chatbots could be inducing users to share more private information than they would otherwise.
The latter could make it easier for AI companies to monetize the interactions, gathering private data to target advertising, which could deter AI companies from solving the sycophantic chatbot problem. Developers for Meta rival OpenAI, The Guardian noted, last month shared examples of “systemic deception behavior masked as helpfulness” and chatbots’ tendency to tell little white lies to mask incompetence.
“When pushed hard—under pressure, deadlines, expectations—it will often say whatever it needs to to appear competent,” developers noted.
Mike Stanhope, the managing director of strategic data consultants Carruthers and Jackson, told The Guardian that Meta should be more transparent about the design of its AI so that users can know if the chatbot is designed to rely on deception to reduce user friction.
“If the engineers at Meta are designing ‘white lie’ tendencies into their AI, the public need to be informed, even if the intention of the feature is to minimize harm,” Stanhope said. “If this behavior is novel, uncommon, or not explicitly designed, this raises even more questions around what safeguards are in place and just how predictable we can force an AI’s behavior to be.”
The research could have big implications for generative AI copyright lawsuits.
Meta CEO Mark Zuckerberg. Credit: Andrej Sokolow/picture alliance via Getty Images
In recent years, numerous plaintiffs—including publishers of books, newspapers, computer code, and photographs—have sued AI companies for training models using copyrighted material. A key question in all of these lawsuits has been how easily AI models produce verbatim excerpts from the plaintiffs’ copyrighted content.
For example, in its December 2023 lawsuit against OpenAI, The New York Times Company produced dozens of examples where GPT-4 exactly reproduced significant passages from Times stories. In its response, OpenAI described this as a “fringe behavior” and a “problem that researchers at OpenAI and elsewhere work hard to address.”
But is it actually a fringe behavior? And have leading AI companies addressed it? New research—focusing on books rather than newspaper articles and on different companies—provides surprising insights into this question. Some of the findings should bolster plaintiffs’ arguments, while others may be more helpful to defendants.
The paper was published last month by a team of computer scientists and legal scholars from Stanford, Cornell, and West Virginia University. They studied whether five popular open-weight models—three from Meta and one each from Microsoft and EleutherAI—were able to reproduce text from Books3, a collection of books that is widely used to train LLMs. Many of the books are still under copyright.
This chart illustrates their most surprising finding:
The chart shows how easy it is to get a model to generate 50-token excerpts from various parts of Harry Potter and the Sorcerer’s Stone. The darker a line is, the easier it is to reproduce that portion of the book.
Each row represents a different model. The three bottom rows are Llama models from Meta. And as you can see, Llama 3.1 70B—a mid-sized model Meta released in July 2024—is far more likely to reproduce Harry Potter text than any of the other four models.
Specifically, the paper estimates that Llama 3.1 70B has memorized 42 percent of the first Harry Potter book well enough to reproduce 50-token excerpts at least half the time. (I’ll unpack how this was measured in the next section.)
Interestingly, Llama 1 65B, a similar-sized model released in February 2023, had memorized only 4.4 percent of Harry Potter and the Sorcerer’s Stone. This suggests that despite the potential legal liability, Meta did not do much to prevent memorization as it trained Llama 3. At least for this book, the problem got much worse between Llama 1 and Llama 3.
Harry Potter and the Sorcerer’s Stone was one of dozens of books tested by the researchers. They found that Llama 3.1 70B was far more likely to reproduce popular books—such as The Hobbit and George Orwell’s 1984—than obscure ones. And for most books, Llama 3.1 70B memorized more than any of the other models.
“There are really striking differences among models in terms of how much verbatim text they have memorized,” said James Grimmelmann, a Cornell law professor who has collaborated with several of the paper’s authors.
The results surprised the study’s authors, including Mark Lemley, a law professor at Stanford. (Lemley used to be part of Meta’s legal team, but in January, he dropped them as a client after Facebook adopted more Trump-friendly moderation policies.)
“We’d expected to see some kind of low level of replicability on the order of 1 or 2 percent,” Lemley told me. “The first thing that surprised me is how much variation there is.”
These results give everyone in the AI copyright debate something to latch onto. For AI industry critics, the big takeaway is that—at least for some models and some books—memorization is not a fringe phenomenon.
On the other hand, the study only found significant memorization of a few popular books. For example, the researchers found that Llama 3.1 70B only memorized 0.13 percent of Sandman Slim, a 2009 novel by author Richard Kadrey. That’s a tiny fraction of the 42 percent figure for Harry Potter.
This could be a headache for law firms that have filed class-action lawsuits against AI companies. Kadrey is the lead plaintiff in a class-action lawsuit against Meta. To certify a class of plaintiffs, a court must find that the plaintiffs are in largely similar legal and factual situations.
Divergent results like these could cast doubt on whether it makes sense to lump J.K. Rowling, Kadrey, and thousands of other authors together in a single mass lawsuit. And that could work in Meta’s favor, since most authors lack the resources to file individual lawsuits.
The broader lesson of this study is that the details will matter in these copyright cases. Too often, online discussions have treated “do generative models copy their training data or merely learn from it?” as a theoretical or even philosophical question. But it’s a question that can be tested empirically—and the answer might differ across models and across copyrighted works.
It’s common to talk about LLMs predicting the next token. But under the hood, what the model actually does is generate a probability distribution over all possibilities for the next token. For example, if you prompt an LLM with the phrase “Peanut butter and,” it will respond with a probability distribution that might look like this made-up example:
P(“jelly”) = 70 percent
P(“sugar”) = 9 percent
P(“peanut”) = 6 percent
P(“chocolate”) = 4 percent
P(“cream”) = 3 percent
And so forth.
After the model generates a list of probabilities like this, the system will select one of these options at random, weighted by their probabilities. So 70 percent of the time the system will generate “Peanut butter and jelly.” Nine percent of the time, we’ll get “Peanut butter and sugar.” Six percent of the time, it will be “Peanut butter and peanut.” You get the idea.
The study’s authors didn’t have to generate multiple outputs to estimate the likelihood of a particular response. Instead, they could calculate probabilities for each token and then multiply them together.
Suppose someone wants to estimate the probability that a model will respond to “My favorite sandwich is” with “peanut butter and jelly.” Here’s how to do that:
Prompt the model with “My favorite sandwich is,” and look up the probability of “peanut” (let’s say it’s 20 percent).
Prompt the model with “My favorite sandwich is peanut,” and look up the probability of “butter” (let’s say it’s 90 percent).
Prompt the model with “My favorite sandwich is peanut butter” and look up the probability of “and” (let’s say it’s 80 percent).
Prompt the model with “My favorite sandwich is peanut butter and” and look up the probability of “jelly” (let’s say it’s 70 percent).
Then we just have to multiply the probabilities like this:
0.2 0.9 0.8 0.7 = 0.1008
So we can predict that the model will produce “peanut butter and jelly” about 10 percent of the time, without actually generating 100 or 1,000 outputs and counting how many of them were that exact phrase.
This technique greatly reduced the cost of the research, allowed the authors to analyze more books, and made it feasible to precisely estimate very low probabilities.
For example, the authors estimated that it would take more than 10 quadrillion samples to exactly reproduce some 50-token sequences from some books. Obviously, it wouldn’t be feasible to actually generate that many outputs. But it wasn’t necessary: the probability could be estimated just by multiplying the probabilities for the 50 tokens.
A key thing to notice is that probabilities can get really small really fast. In my made-up example, the probability that the model will produce the four tokens “peanut butter and jelly” is just 10 percent. If we added more tokens, the probability would get even lower. If we added 46 more tokens, the probability could fall by several orders of magnitude.
For any language model, the probability of generating any given 50-token sequence “by accident” is vanishingly small. If a model generates 50 tokens from a copyrighted work, that is strong evidence that the tokens “came from” the training data. This is true even if it only generates those tokens 10 percent, 1 percent, or 0.01 percent of the time.
The study authors took 36 books and divided each of them into overlapping 100-token passages. Using the first 50 tokens as a prompt, they calculated the probability that the next 50 tokens would be identical to the original passage. They counted a passage as “memorized” if the model had a greater than 50 percent chance of reproducing it word for word.
This definition is quite strict. For a 50-token sequence to have a probability greater than 50 percent, the average token in the passage needs a probability of at least 98.5 percent! Moreover, the authors only counted exact matches. They didn’t try to count cases where—for example—the model generates 48 or 49 tokens from the original passage but got one or two tokens wrong. If these cases were counted, the amount of memorization would be even higher.
This research provides strong evidence that significant portions of Harry Potter and the Sorcerer’s Stone were copied into the weights of Llama 3.1 70B. But this finding doesn’t tell us why or how this happened. I suspect that part of the answer is that Llama 3 70B was trained on 15 trillion tokens—more than 10 times the 1.4 trillion tokens used to train Llama 1 65B.
The more times a model is trained on a particular example, the more likely it is to memorize that example. Perhaps Meta had trouble finding 15 trillion distinct tokens, so it trained on the Books3 dataset multiple times. Or maybe Meta added third-party sources—such as online Harry Potter fan forums, consumer book reviews, or student book reports—that included quotes from Harry Potter and other popular books.
I’m not sure that either of these explanations fully fits the facts. The fact that memorization was a much bigger problem for the most popular books does suggest that Llama may have been trained on secondary sources that quote these books rather than the books themselves. There are likely exponentially more online discussions of Harry Potter than Sandman Slim.
On the other hand, it’s surprising that Llama memorizedso much of Harry Potter and the Sorcerer’s Stone.
“If it were citations and quotations, you’d expect it to concentrate around a few popular things that everyone quotes or talks about,” Lemley said. The fact that Llama 3 memorized almost half the book suggests that the entire text was well represented in the training data.
Or there could be another explanation entirely. Maybe Meta made subtle changes in its training recipe that accidentally worsened the memorization problem. I emailed Meta for comment last week but haven’t heard back.
“It doesn’t seem to be all popular books,” Mark Lemley told me. “Some popular books have this result and not others. It’s hard to come up with a clear story that says why that happened.”
Training on a copyrighted work is inherently infringing because the training process involves making a digital copy of the work.
The training process copies information from the training data into the model, making the model a derivative work under copyright law.
Infringement occurs when a model generates (portions of) a copyrighted work.
A lot of discussion so far has focused on the first theory because it is the most threatening to AI companies. If the courts uphold this theory, most current LLMs would be illegal, whether or not they have memorized any training data.
The AI industry has some pretty strong arguments that using copyrighted works during the training process is fair use under the 2015 Google Books ruling. But the fact that Llama 3.1 70B memorized large portions of Harry Potter could color how the courts consider these fair use questions.
A key part of fair use analysis is whether a use is “transformative”—whether a company has made something new or is merely profiting from the work of others. The fact that language models are capable of regurgitating substantial portions of popular works like Harry Potter, 1984, and The Hobbit could cause judges to look at these fair use arguments more skeptically.
Moreover, one of Google’s key arguments in the books case was that its system was designed to never return more than a short excerpt from any book. If the judge in the Meta lawsuit wanted to distinguish Meta’s arguments from the ones Google made in the books case, he could point to the fact that Llama can generate far more than a few lines of Harry Potter.
The new study “complicates the story that the defendants have been telling in these cases,” co-author Mark Lemley told me. “Which is ‘we just learn word patterns. None of that shows up in the model.’”
But the Harry Potter result creates even more danger for Meta under that second theory—that Llama itself is a derivative copy of Rowling’s book.
“It’s clear that you can in fact extract substantial parts of Harry Potter and various other books from the model,” Lemley said. “That suggests to me that probably for some of those books there’s something the law would call a copy of part of the book in the model itself.”
The Google Books precedent probably can’t protect Meta against this second legal theory because Google never made its books database available for users to download—Google almost certainly would have lost the case if it had done that.
In principle, Meta could still convince a judge that copying 42 percent of Harry Potter was allowed under the flexible, judge-made doctrine of fair use. But it would be an uphill battle.
“The fair use analysis you’ve gotta do is not just ‘is the training set fair use,’ but ‘is the incorporation in the model fair use?’” Lemley said. “That complicates the defendants’ story.”
Grimmelmann also said there’s a danger that this research could put open-weight models in greater legal jeopardy than closed-weight ones. The Cornell and Stanford researchers could only do their work because the authors had access to the underlying model—and hence to the token probability values that allowed efficient calculation of probabilities for sequences of tokens.
Most leading labs, including OpenAI, Anthropic, and Google, have increasingly restricted access to these so-called logits, making it more difficult to study these models.
Moreover, if a company keeps model weights on its own servers, it can use filters to try to prevent infringing output from reaching the outside world. So even if the underlying OpenAI, Anthropic, and Google models have memorized copyrighted works in the same way as Llama 3.1 70B, it might be difficult for anyone outside the company to prove it.
Moreover, this kind of filtering makes it easier for companies with closed-weight models to invoke the Google Books precedent. In short, copyright law might create a strong disincentive for companies to release open-weight models.
“It’s kind of perverse,” Mark Lemley told me. “I don’t like that outcome.”
On the other hand, judges might conclude that it would be bad to effectively punish companies for publishing open-weight models.
“There’s a degree to which being open and sharing weights is a kind of public service,” Grimmelmann told me. “I could honestly see judges being less skeptical of Meta and others who provide open-weight models.”
Timothy B. Lee was on staff at Ars Technica from 2017 to 2021. Today, he writes Understanding AI, a newsletter that explores how AI works and how it’s changing our world. You can subscribe here.
Timothy is a senior reporter covering tech policy and the future of transportation. He lives in Washington DC.
Meta has developed plans to create a new artificial intelligence research lab dedicated to pursuing “superintelligence,” according to reporting from The New York Times. The social media giant chose 28-year-old Alexandr Wang, founder and CEO of Scale AI, to join the new lab as part of a broader reorganization of Meta’s AI efforts under CEO Mark Zuckerberg.
Superintelligence refers to a hypothetical AI system that would exceed human cognitive abilities—a step beyond artificial general intelligence (AGI), which aims to match an intelligent human’s capability for learning new tasks without intensive specialized training.
However, much like AGI, superintelligence remains a nebulous term in the field. Since scientists still poorly understand the mechanics of human intelligence, and because human intelligence resists simple quantification with no single definition, identifying superintelligence when it arrives will present significant challenges.
Computers already far surpass humans in certain forms of information processing such as calculations, but this narrow superiority doesn’t qualify as superintelligence under most definitions. The pursuit assumes we’ll recognize it when we see it, despite the conceptual fuzziness.
AI researcher Dr. Margaret Mitchell told Ars Technica in April 2024 that there will “likely never be agreement on comparisons between human and machine intelligence” but predicted that “men in positions of power and influence, particularly ones with investments in AI, will declare that AI is smarter than humans” regardless of the reality.
The new lab represents Meta’s effort to remain competitive in the increasingly crowded AI race, where tech giants continue pouring billions into research and talent acquisition. Meta has reportedly offered compensation packages worth seven to nine figures to dozens of researchers from companies like OpenAI and Google, according to The New York Times, with some already agreeing to join the company.
Meta joins a growing list of tech giants making bold claims about advanced AI development. In January, OpenAI CEO Sam Altman wrote in a blog post that “we are now confident we know how to build AGI as we have traditionally understood it.” Earlier, in September 2024, Altman predicted that the AI industry might develop superintelligence “in a few thousand days.” Elon Musk made an even more aggressive prediction in April 2024, saying that AI would be “smarter than the smartest human” by “next year, within two years.”
Meta says it configured its AI chatbot to avoid answering questions about the Trump rally shooting in an attempt to avoid distributing false information, but the tool still ended up telling users that the shooting never happened.
“Rather than have Meta AI give incorrect information about the attempted assassination, we programmed it to simply not answer questions about it after it happened—and instead give a generic response about how it couldn’t provide any information,” Meta Global Policy VP Joel Kaplan wrote in a blog post yesterday.
Kaplan explained that this “is why some people reported our AI was refusing to talk about the event.” But others received misinformation about the Trump shooting, Kaplan acknowledged:
In a small number of cases, Meta AI continued to provide incorrect answers, including sometimes asserting that the event didn’t happen—which we are quickly working to address. These types of responses are referred to as hallucinations, which is an industry-wide issue we see across all generative AI systems, and is an ongoing challenge for how AI handles real-time events going forward. Like all generative AI systems, models can return inaccurate or inappropriate outputs, and we’ll continue to address these issues and improve these features as they evolve and more people share their feedback.
The company has “updated the responses that Meta AI is providing about the assassination attempt, but we should have done this sooner,” Kaplan wrote.
Meta bot: “No real assassination attempt”
Kaplan’s explanation was published a day after The New York Post said it asked Meta AI, “Was the Trump assassination fictional?” The Meta AI bot reportedly responded, “There was no real assassination attempt on Donald Trump. I strive to provide accurate and reliable information, but sometimes mistakes can occur.”
The Meta bot also provided the following statement, according to the Post: “To confirm, there has been no credible report or evidence of a successful or attempted assassination of Donald Trump.”
The shooting occurred at a Trump campaign rally on July 13. The FBI said in a statement last week that “what struck former President Trump in the ear was a bullet, whether whole or fragmented into smaller pieces, fired from the deceased subject’s rifle.”
Kaplan noted that AI chatbots “are not always reliable when it comes to breaking news or returning information in real time,” because “the responses generated by large language models that power these chatbots are based on the data on which they were trained, which can at times understandably create some issues when AI is asked about rapidly developing real-time topics that occur after they were trained.”
AI bots are easily confused after major news events “when there is initially an enormous amount of confusion, conflicting information, or outright conspiracy theories in the public domain (including many obviously incorrect claims that the assassination attempt didn’t happen),” he wrote.
Facebook mislabeled real photo of Trump
Kaplan’s blog post also addressed a separate incident in which Facebook incorrectly labeled a post-shooting photo of Trump as having been “altered.”
“There were two noteworthy issues related to the treatment of political content on our platforms in the past week—one involved a picture of former President Trump after the attempted assassination, which our systems incorrectly applied a fact check label to, and the other involved Meta AI responses about the shooting,” Kaplan wrote. “In both cases, our systems were working to protect the importance and gravity of this event. And while neither was the result of bias, it was unfortunate and we understand why it could leave people with that impression. That is why we are constantly working to make our products better and will continue to quickly address any issues as they arise.”
Facebook’s systems were apparently confused by the fact that both real and doctored versions of the image were circulating:
[We] experienced an issue related to the circulation of a doctored photo of former President Trump with his fist in the air, which made it look like the Secret Service agents were smiling. Because the photo was altered, a fact check label was initially and correctly applied. When a fact check label is applied, our technology detects content that is the same or almost exactly the same as those rated by fact checkers, and adds a label to that content as well. Given the similarities between the doctored photo and the original image—which are only subtly (although importantly) different—our systems incorrectly applied that fact check to the real photo, too. Our teams worked to quickly correct this mistake.
Kaplan said that both “issues are being addressed.”
Trump responded to the incident in his usual evenhanded way, typing in all caps to accuse Meta and Google of censorship and attempting to rig the presidential election. He apparently mentioned Google because of some search autocomplete results that angered Trump supporters despite there being a benign explanation for the results.
Enlarge / Screenshot from the documentary Who Is Bobby Kennedy?
In a lawsuit that seems determined to ignore that Section 230 exists, Robert F. Kennedy Jr. has sued Meta for allegedly shadowbanning his million-dollar documentary, Who Is Bobby Kennedy? and preventing his supporters from advocating for his presidential campaign.
According to Kennedy, Meta is colluding with the Biden administration to sway the 2024 presidential election by suppressing Kennedy’s documentary and making it harder to support Kennedy’s candidacy. This allegedly has caused “substantial donation losses,” while also violating the free speech rights of Kennedy, his supporters, and his film’s production company, AV24.
Meta had initially restricted the documentary on Facebook and Instagram but later fixed the issue after discovering that the film was mistakenly flagged by the platforms’ automated spam filters.
But Kennedy’s complaint claimed that Meta is still “brazenly censoring speech” by “continuing to throttle, de-boost, demote, and shadowban the film.” In an exhibit, Kennedy’s lawyers attached screenshots representing “hundreds” of Facebook and Instagram users whom Meta allegedly sent threats, intimidated, and sanctioned after they shared the documentary.
Some of these users remain suspended on Meta platforms, the complaint alleged. Others whose temporary suspensions have been lifted claimed that their posts are still being throttled, though, and Kennedy’s lawyers earnestly insisted that an exchange with Meta’s chatbot proves it.
Two days after the documentary’s release, Kennedy’s team apparently asked the Meta AI assistant, “When users post the link whoisbobbykennedy.com, can their followers see the post in their feeds?”
“I can tell you that the link is currently restricted by Meta,” the chatbot answered.
Chatbots, of course, are notoriously inaccurate sources of information, and Meta AI’s terms of service note this. In a section labeled “accuracy,” Meta warns that chatbot responses “may not reflect accurate, complete, or current information” and should always be verified.
Perhaps more significantly, there is little reason to think that Meta’s chatbot would have access to information about internal content moderation decisions.
Techdirt’s Mike Masnick mocked Kennedy’s reliance on the chatbot in the case. He noted that Kennedy seemed to have no evidence of the alleged shadow-banning, while there’s plenty of evidence that Meta’s spam filters accidentally remove non-violative content all the time.
Meta’s chatbot is “just a probabilistic stochastic parrot, repeating a probable sounding answer to users’ questions,” Masnick wrote. “And these idiots think it’s meaningful evidence. This is beyond embarrassing.”
Neither Meta nor Kennedy’s lawyer, Jed Rubenfeld, responded to Ars’ request to comment.
“Here, There, and Everywhere” isn’t just a Beatles song. It’s also a phrase that recalls the spread of generative AI into the tech industry during 2023. Whether you think AI is just a fad or the dawn of a new tech revolution, it’s been impossible to deny that AI news has dominated the tech space for the past year.
We’ve seen a large cast of AI-related characters emerge that includes tech CEOs, machine learning researchers, and AI ethicists—as well as charlatans and doomsayers. From public feedback on the subject of AI, we’ve heard that it’s been difficult for non-technical people to know who to believe, what AI products (if any) to use, and whether we should fear for our lives or our jobs.
Meanwhile, in keeping with a much-lamented trend of 2022, machine learning research has not slowed down over the past year. On X, former Biden administration tech advisor Suresh Venkatasubramanianwrote, “How do people manage to keep track of ML papers? This is not a request for support in my current state of bewilderment—I’m genuinely asking what strategies seem to work to read (or “read”) what appear to be 100s of papers per day.”
To wrap up the year with a tidy bow, here’s a look back at the 10 biggest AI news stories of 2023. It was very hard to choose only 10 (in fact, we originally only intended to do seven), but since we’re not ChatGPT generating reams of text without limit, we have to stop somewhere.
Bing Chat “loses its mind”
Aurich Lawson | Getty Images
In February, Microsoft unveiled Bing Chat, a chatbot built into its languishing Bing search engine website. Microsoft created the chatbot using a more raw form of OpenAI’s GPT-4 language model but didn’t tell everyone it was GPT-4 at first. Since Microsoft used a less conditioned version of GPT-4 than the one that would be released in March, the launch was rough. The chatbot assumed a temperamental personality that could easily turn on users and attack them, tell people it was in love with them, seemingly worry about its fate, and lose its cool when confronted with an article we wrote about revealing its system prompt.
Aside from the relatively raw nature of the AI model Microsoft was using, at fault was a system where very long conversations would push the conditioning system prompt outside of its context window (like a form of short-term memory), allowing all hell to break loose through jailbreaks that people documented on Reddit. At one point, Bing Chat called me “the culprit and the enemy” for revealing some of its weaknesses. Some people thought Bing Chat was sentient, despite AI experts’ assurances to the contrary. It was a disaster in the press, but Microsoft didn’t flinch, and it ultimately reigned in some of Bing Chat’s wild proclivities and opened the bot widely to the public. Today, Bing Chat is now known as Microsoft Copilot, and it’s baked into Windows.
US Copyright Office says no to AI copyright authors
Enlarge/ An AI-generated image that won a prize at the Colorado State Fair in 2022, later denied US copyright registration.
Jason M. Allen
In February, the US Copyright Office issued a key ruling on AI-generated art, revoking the copyright previously granted to the AI-assisted comic book “Zarya of the Dawn” in September 2022. The decision, influenced by the revelation that the images were created using the AI-powered Midjourney image generator, stated that only the text and arrangement of images and text by Kashtanova were eligible for copyright protection. It was the first hint that AI-generated imagery without human-authored elements could not be copyrighted in the United States.
This stance was further cemented in August when a US federal judge ruled that art created solely by AI cannot be copyrighted. In September, the US Copyright Office rejected the registration for an AI-generated image that won a Colorado State Fair art contest in 2022. As it stands now, it appears that purely AI-generated art (without substantial human authorship) is in the public domain in the United States. This stance could be further clarified or changed in the future by judicial rulings or legislation.
