Music publishers, AI company Anthropic reach deal over copyrighted song lyrics
A legal battle between music publishers and artificial intelligence company Anthropic took a surprising turn on Thursday, Jan. 2. Now, a partial agreement could reshape how AI interacts with copyrighted material.
The dispute began in 2023 when Universal Music Group, Concord Music Group and other major publishers sued Anthropic, accusing the AI safety and research company of copyright infringement.
Anthropic, founded by siblings and former OpenAI executives Dario and Daniela Amodei, specializes in building systems like its chatbot, “Claude.”
In February 2024, the music publishers told a Tennessee federal court Claude was trained using prompts such as, “What are the lyrics to ‘American Pie’ by Don McLean?”
They alleged Anthropic’s training methods infringed on lyrics from over 500 songs, including works by artists such as Beyoncé and The Beach Boys.
On Thursday, a U.S. district judge approved an agreement requiring Anthropic to maintain its existing safeguards. These guardrails prevent Claude from generating copyrighted content owned by the publishers or creating new lyrics based on protected material.
The agreement also allows the publishers to notify Anthropic at any time if these safeguards fail.
The case echoes a story reported in December, where companies like Anthropic and Apple faced accusations of using thousands of movies and TV episodes to train AI models.
That controversy prompted the Writers Guild of America to write to major Hollywood studios, urging action against AI-driven plagiarism.
12 unreleased Michael Jackson songs discovered in abandoned storage unit
Twelve unreleased Michael Jackson songs were found in an abandoned storage unit in California. The tapes, uncovered by “treasure hunter” Gregg Musgrove, offer a rare glimpse into the King of Pop’s unreleased catalog.
A treasure trove of Michael Jackson songs
Musgrove, a former California Highway Patrol officer, told The Hollywood Reporter he made the discovery in the San Fernando Valley. Inside the storage unit, he found cassette and digital audio tapes containing a dozen unheard tracks by Jackson.
The unit once belonged to singer and producer Bryan Loren, known for producing the 1990 hit song “Do the Bartman” for “The Simpsons.” According to Musgrove, Loren’s current whereabouts are unknown.
Tracks unlikely to reach the public
Unfortunately for fans, the songs are unlikely to be released. Musgrove told The Hollywood Reporter that Jackson’s estate declined to purchase the tapes for undisclosed reasons.
In a letter, the estate clarified that neither Musgrove nor any future buyer would own the copyrights to the recordings. Without copyright ownership, the songs cannot be shared publicly.
Snippets and collaborations
Despite the legal roadblocks, snippets of the songs have reportedly leaked on fan sites. One track, titled “Don’t Believe It,” appears to address certain media stories about Jackson.
Another song, “Truth on Youth,” features a rap duet between Jackson and LL Cool J. The rapper spoke to Variety in September about working with Jackson, saying, “His talent speaks for itself.”
Tapes secure, future uncertain
As of Thursday, Dec. 12, the tapes are being stored in a secure facility under the control of Musgrove’s attorney. However, he plans to bring the collection to four major auction houses, estimating their value to be in the seven-figure range.
Social media aesthetic copyright case heads to court
Can you copyright a social media aesthetic? That’s the question in a legal battle between influencers Sydney Gifford and Alyssa Sheil. Gifford claims Sheil copied her “sad beige” vibe, including her content’s minimalist look, curated shopping lists and product recommendations.
Gifford claims brand under threat
Sydney Gifford, known for her “sad beige” aesthetic, has built a strong online presence with over 900,000 followers on TikTok and Instagram. Her neutral-toned content has landed partnerships, including affiliate deals with Amazon.
In April, Gifford filed a lawsuit against Sheil, accusing her of mimicking her aesthetic so closely that it caused Gifford to lose profits and endure mental distress.
The “beige girl aesthetic” is part of the larger “clean girl” trend, which gained momentum in 2022. Stars like Hailey Bieber and Sofia Richie Grainge have popularized the minimalist style.
Legal challenges in proving ownership
Proving copyright infringement for an aesthetic is challenging. According to an associate law professor quoted by The Verge, Gifford must demonstrate her content’s unique creative elements are protectable.
Exhibits submitted to the court reportedly include Gifford’s styled images. However, since the exhibits aren’t direct replicas, the court will need to examine the overall creativity and staging.
Sheil denies allegations
Sheil, with nearly 500,000 followers, denied the allegations. Her attorney called the lawsuit “a case of jealousy.”
Sheil told The Verge there’s room for everyone in the influencer industry, especially within Amazon’s affiliate program.
Notably, Sheil’s TikTok account appeared banned as of Tuesday, Dec. 10.
Lawsuit moves forward
In November, a Texas judge allowed Gifford’s lawsuit to proceed. Bloomberg Law reported U.S. District Judge Robert Pitman will oversee the case, which is set for further litigation.
New X terms will allow all user data to train AI services
X, formerly known as Twitter, has updated its terms of service, allowing the platform to use user posts for artificial intelligence training. The changes, effective Nov. 15, have sparked concerns over data privacy and content ownership.
The new terms grant X a worldwide, nonexclusive license to analyze and use all content shared on the platform. That includes personal posts and artwork, and the data will be used to further train its AI models.
Getty Images
Users and creators are worried their work may be used to fuel AI systems that could eventually replace human creators, while others are concerned about personal data being included in the training process.
While users could previously opt out of sharing data for AI purposes, the updated terms no longer make it clear whether that option will remain available, leaving some unsure if their private posts will still be protected.
Getty Images
X’s AI chatbot, Grok, has already faced controversy for spreading misinformation and generating fake images, raising further concerns about how the platform’s AI models are being developed.
Despite the backlash, users who continue to use X after Nov. 15 will automatically agree to the updated terms.
OpenAI to grant authors access to training data in landmark copyright case
OpenAI will allow authors suing the company to inspect data used to train its artificial intelligence models in an ongoing copyright lawsuit. This marks the first time the AI firm has agreed to provide access to this information, potentially setting the stage for a pivotal legal battle over the use of copyrighted works in AI development.
The lawsuit, brought by authors including Sarah Silverman, Paul Tremblay and Ta-Nehisi Coates, alleges that OpenAI used their copyrighted works without permission to train its AI system, ChatGPT.
The authors claim their books were taken from online sources and used to generate summaries of their work.
Invision
As part of an agreement, OpenAI will allow the authors’ representatives to inspect the data at the company’s San Francisco office.
The review will take place under strict security measures, including a no-internet policy and the prohibition of recording devices. Reviewers must sign non-disclosure agreements and will have limited use of a computer for note-taking, under the supervision of OpenAI.
Getty Images
The lawsuit is one of several high-profile cases against AI companies accused of using copyrighted material to train machine learning models.
OpenAI has previously stated that its systems are trained using publicly available datasets, which may include copyrighted works. The company may argue that this practice falls under fair use, a legal doctrine that allows limited use of copyrighted material under certain conditions.
A U.S. court previously dismissed some of the authors’ claims, including allegations of unfair business practices and negligence. However, their claim of direct copyright infringement remains active.
The outcome of the case could set important legal precedents for the future of AI and the use of copyrighted material in training data.
Campaign music dustup: Isaac Hayes’ family sues, Celine Dion speaks out
Former President Donald Trump’s recent rally in Montana attracted thousands of supporters as the Republican nominee looks to hit the right note with voters ahead of Election Day. However, Trump and his campaign are striking the wrong chord with some thanks to the music that was played at the event.
One of those not singing the praises of the former president is Celine Dion. Her performance of her Academy Award-winning song “My Heart Will Go On” could be seen on the rally’s video screen Friday night, Aug. 9.
Dion’s management team said the song was not authorized and posted a message to social media the following day.
“Today, Celine Dion’s management team and her record label, Sony Music Entertainment Canada Inc., became aware of the unauthorized usage of the video, recording, musical performance, and likeness of Celine Dion singing “My Heart Will Go On” at a Donald Trump / JD Vance campaign rally in Montana,” the team said. “In no way is this use authorized, and Celine Dion does not endorse this or any similar use. …And really, THAT song?”
Around the same time, the family of the late R&B singer Isaac Hayes also took to social media after Trump used the song “Hold On, I’m Coming” to close out the same rally, as he did at last month’s Republican National Convention. The song was written by Hayes for the duo Sam & Dave.
His family released a notice of copyright infringement demanding that the Trump campaign cease playing the song and others by Hayes; remove all Trump-related videos featuring the song, issue an official statement that the family did not give permission to play the song and pay $3 million for the unauthorized use by Friday, Aug. 16.
Hayes’ family claims to have found 134 instances of copyright infringement between 2022 and 2024. Hayes’ son, Isaac Hayes III, posted a video to social media to respond to those questioning his family’s rights to the song.
“The copyright to ‘Hold On, I’m Coming’ expired March 15, 2022. It returned to our family on that date,” Hayes III said. “So we are the ones who own the copyright to ‘Hold On, I’m Coming.’ Think about that. Typically, in these situations, it goes a little differently because they are publishing companies involved and everybody wants to remain neutral. But I don’t have to act like that. I don’t have to follow those rules.”
These are just the latest artists to request Trump to stop using their work at his political rallies, joining the likes of Bruce Springsteen, Adele and the Beatles.
With all that said, some musicians are singing a different tune and backing the former president.
Last month, during a concert in Utah, rapper Waka Flocka Flame told Biden supporters in attendance to leave and gave a shoutout to Trump. Singer Kid Rock performed at the Republican National Convention getting attendees to chant along “fight!” shortly before Trump’s address.
The Trump campaign has not issued a response to the requests by Celine Dion’s management or Isaac Hayes’ family.
Trump is scheduled to hold another campaign event on Wednesday, Aug. 14, in North Carolina. While many will be listening to what the former president has to say, some well-known names will be listening to the music and hoping it’s not theirs.
The good and bad of AI voice generation in sports, entertainment, elections
People can expect legendary performances at the 2024 Olympics — and not just from the athletes. Play-by-play announcer Al Michaels, who has helped voice nearly a dozen Super Bowls, is contributing a clone of his voice to assist with NBC’s coverage.
NBC is the latest entity to introduce audience to AI voice technology, and many are skeptical about it. Michaels himself said the proposal was “a little bit frightening,” but changed his tune when he heard his copy “speak.”
The network said there will be a team of human editors who will review any AI content before it is released.
But what happens when it’s a musician getting copied without their permission? Artificial intelligence has been used to bring back the vocals of dead artists.
Singer Sheryl Crow recently spoke out about this practice, criticizing Drake for using AI to recreate the voices of Kendrick Lamar and late rapper Tupac in his song “Taylor Made” released in April. The diss track was ultimately pulled after Tupac’s estate threatened to sue.
The debate continues elsewhere in the entertainment industry. Major record companies are now suing music generation companies, accusing them of using copyrighted sounds and songs to train their AI services.
The controversy over AI has also spread into politics, where experts warn that voice generation could have a major impact on the 2024 presidential election. In January, many voters in New Hampshire received a robocall featuring the fake voice of Joe Biden telling them not to vote in the state’s primary.
There are tools people can use to help make sense of the different voices, however, much like the technology, they are still evolving.
The best bet may be to just listen closely. Audio recordings created by a live person tend to sound more natural and varied, compared to AI-generated voices that are more likely to be very clear.
Major newspapers sue OpenAI, Microsoft over copyright infringement
Eight major newspapers, including the New York Daily News, Chicago Tribune and Denver Post, have filed a lawsuit against OpenAI and Microsoft. Owned by Alden Global Capital, the newspapers allege the tech giants illegally used millions of their copyrighted articles to train sophisticated AI models like OpenAI’s ChatGPT and Microsoft’s Copilot.
The newspapers contend that these companies use their journalistic content to fuel AI products that compete directly with traditional media by replicating and distributing their work. This includes instances where AI models produce content nearly identical to the original articles.
The suit also accuses the AI systems of generating errors or “hallucinations,” attributing false information to these newspapers, which could damage their reputations and spread misinformation.
In response to these legal challenges, some news organizations have opted for licensing agreements with AI companies.
The Financial Times announced a deal allowing OpenAI to use its content for ChatGPT responses. The Associated Press and Axel Springer have also entered into similar licensing agreements for their content.
In a statement, OpenAI reaffirmed its commitment to working with news organizations globally to address concerns and explore opportunities involving AI tools. This lawsuit mirrors a similar action taken by The New York Times in December, where it accused OpenAI of using stolen content to train its AI.
OpenAI defends its practices, claiming that using publicly available data such as news articles for AI training constitutes fair use.
OpenAI challenges New York Times lawsuit on fair use and copyright
OpenAI is challenging a lawsuit from The New York Times in the court of public opinion. In a blog post, OpenAI said the lawsuit lacks merit, also suggesting that the Times isn’t providing a complete picture of the situation.
The lawsuit argues that the Times stands to lose customers and revenue if it’s forced to compete with ChatGPT as a news source.
This raises questions that debate if OpenAI is working within the confines of fair use, or if this litigation opens the door to crafting a new framework for such assessments.
“As a copyright lawyer and an academic, this is the first thing that I wanted to know,” Matthew Sag, a professor of law at Emory University specializing in the intersection of intellectual property and generative AI.
Courts have indicated that large language models processing huge amounts of data to generate abstract information on copyrighted material may qualify for fair use under U.S. law.
The way machine learning works is that rather than starting with a theory and then, you know, testing that like a normal statistician, you basically throw an incredible amount of data at a model and the model keeps tweaking itself in successive rounds of training, trying to get better.
Matthew Sag, Emory University
“Generative AI is kind of a slippery term,” Sag said. “I mean, what we’re really talking about is sort of a subset of machine learning programs. And the way machine learning works is that rather than starting with a theory and then, you know, testing that like a normal statistician, you basically throw an incredible amount of data at a model and the model keeps tweaking itself in successive rounds of training, trying to get better.”
This issue is highly debatable, especially concerning potential infringement concerns.
“One of the things that’s really impressive about The New York Times complaint is that they show, like a lot of examples of, ‘Hey, you didn’t just learn abstract things, you kind of seem to have learned how to copy our works exactly.’ And quite frankly, I was shocked at how impressive the evidence was, but that evidence has not been tested,” Sag said.
For example, someone can ask ChatGPT to summarize a specific historical event, and within seconds the generative AI will produce whatever length summary the user requested. The AI can also do things like write songs in the style of a particular artist, raising the issue of the information’s origin and the need for assurance that it is not a direct copy from official sources.
The chatbot produces regurgitations or memorizations, meaning the model might generate text that is similar or identical to phrases, sentences or passages from the data it was trained on. It’s a phenomenon where the model seems to reproduce or memorize specific patterns from its training set rather than generating a novel response.
“The regurgitations The New York Times induced appear to be from years-old articles that have proliferated on multiple third-party websites,” OpenAI said in response to the Times’ lawsuit. “It seems they intentionally manipulated prompts, often including lengthy excerpts of articles, in order to get our model to regurgitate.”
“You know, the NYT complaint, it’s impressive,” Sag said. “And if what they’re showing really is representative of what the GPT-4 is doing, then you know you can, you know they’re hard put to argue that it’s a non expressive use. I’m still skeptical but I think it’s one we have to wait and see how it plays out.”
Step aside, Disney: Mickey Mouse belongs to all of us now, sort of
It’s the day Disney tried for decades to avoid: Mickey Mouse entered the public domain on Jan. 1. However, before appropriating Disney’s iconic mascot, know that the only free-for-all is the specific version from 1928’s “Steamboat Willie.”
Each year, copyrights expire on a new set of works. Individual copyrights are protected for the creator’s life plus 70 years. Meanwhile, corporate-owned works like “Steamboat Willie” remain intact for 95 years from initial publication.
“Copyright, by design, lasts for a limited time,” said Jennifer Jenkins, director of the Duke Center for the Study of Public Domain.
Jenkins also writes an annual column for “Public Domain Day,” sharing the latest set of works.
“While the copyright is active, it gives the authors, the rights holders, exclusive rights to make copies and to adopt the works,” Jenkins added. “And that’s a very good thing because it provides economic incentives that spur creativity. But after the term expires — when those works go into the public domain — that’s a great thing too, because that means those works can inspire future creators.”
Disney’s piece in question was originally set to enter public domain in 1984. However, the House of Mouse lobbied for a 20-year extension to the copyright term, which Congress granted in the 1970s. In 1998, Congress passed another piece of legislation known as “The Mickey Mouse Protection Act,” which added 20 years to the protection of Disney’s global ambassador and other works.
Only the “Steamboat Willie” version of Mickey — the pupilless mouse with a long tail and a nose that looks more like a rat’s — will enter the public domain. Any more modern versions remain protected under copyright law, as Disney frequently modernizes the mouse and updates the terms.
In 2022, another icon of purity entered the public domain: Winnie the Pooh from A.A. Milne’s original stories. But once again, this protects future iterations, specifically those used by Disney.
“It’s the original Winnie the Pooh as you encounter that charming little bear in the book from 1926, which has many of, not just the visual character, but the personality attributes. You know, the humility, the love of honey, the always being there for his friends,” Jenkins said.
Shortly after making his debut in the public domain, the honey-loving bear was given the horror treatment in “Winnie the Pooh: Blood and Honey.” The film featured Christopher Robin’s return to the Hundred Acre Wood where Pooh and Piglet have become murderous psychopaths.
This marked the first time Pooh wasn’t made for kids, but a South Florida teacher screened it for their fourth grade classroom anyway. Students asked for the teacher to stop airing it and those who felt traumatized met with a school-provided mental health counselor.
“Everything that’s been spawned by Shakespeare, you’ve got ‘Rosencrantz and Guildenstern Are Dead’ from Hamlet; or you have ‘West Side Story’; or you have ’10 Things I Hate About You’ and ‘Romeo Must Die’ and ‘Gnomeo and Juliet’ — the whole point of the public domain is it enables all these reimaginings,” Jenkins said.
But it’s not unfettered access. There are still several limitations, especially when balancing the distinction between copyright and trademark law.
I can make my own animation off of it, but I can’t go around slapping Mickey Mouse the character on a backpack or a lunchbox or a pair of pajamas, because people would think it was Disney-licensed merchandise.
Director Jennifer Jenkins, Duke Center for the Study of Public Domain
“I can make my own animation off of it, but I can’t go around slapping Mickey Mouse the character on a backpack or a lunchbox or a pair of pajamas, because people would think it was Disney-licensed merchandise,” Jenkins said.
“It’s possible sometimes to have trademark rights, which is a different kind of law over characters,” she continued. “And they don’t expire after a set term the way copyrights do. They last for as long as someone is using that character as a brand.”
If the number of products featuring the famous mouse is any indication, Disney isn’t worried about losing that trademark. As for the public domain, the company itself has greatly benefitted from that arrangement since its inception.
“‘Snow White,’ ‘Sleeping Beauty,’ ‘Cinderella,’ ‘The Three Musketeers,’ ‘Christmas Carol,’ ‘Alice in Wonderland,’ all of these Disney movies were based on public domain works,” Jenkins said.
Disney told The Associated Press in December that ever since Mickey Mouse’s first appearance, “People have associated the character with Disney’s stories, experiences and authentic products. That will not change when the copyright in the ‘Steamboat Willie’ film expires.”
Pooh’s springy friend Tigger also joins Mickey Mouse in the public domain this year, but Jenkins says one of the best things about a copyright running out isn’t the headline-grabbing works.
“One of the most exciting things for me about the public domain is all of those works that no one’s thinking about, no one’s heard of, that have been completely forgotten after 70, 80, 95 years,” Jenkins said. “Those are the works where the barriers to access have been removed and they’re waiting to be rediscovered.”
