Meta's Llama 3.1 AI model reproduces copyrighted book text, raising legal alarms
Among five major AI models analyzed for how they processed the Books3 dataset, Llama 3.1—Meta's 70-billion parameter model released in July 2024—was the most prone to output copyrighted content. By contrast, Llama 1 65B, released in February 2023, had memorized only 4.4% of the same book, highlighting a significant increase in verbatim retention over time.
The model was also found to reproduce exact excerpts from other iconic works such as The Hobbit and 1984. Experts suspect this could be due to repeated exposure to the same texts during training, possibly sourced from fan sites, academic analyses, or online reviews. Adjustments to Meta's training strategy may have unintentionally worsened the memorization problem.
These findings come amid growing legal pressure on AI developers. The New York Times has already filed a lawsuit against OpenAI and Microsoft, accusing them of copyright infringement by training models like ChatGPT on proprietary articles. The lawsuit claims OpenAI's models can not only reproduce content verbatim but also mimic The Times' unique style.
For Meta, these revelations may trigger similar legal risks as calls for transparency and ethical AI development intensify across the industry.
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Mint
2 hours ago
- Mint
SoftBank's $47 Billion AI-Led Stock Rally Is at Risk of Stalling
SoftBank Group Corp. shares look to have limited upside after the rally this month that added more than ¥7 trillion to an all-time high Monday. Founder Masayoshi Son's aggressive investment in artificial intelligence, from its $500 billion data center project with OpenAI and Oracle Corp., to growing holdings in Nvidia Corp., drove stellar gains in the stock. However it failed to keep its momentum Tuesday, despite plans to invest $2 billion in Intel Corp., with investors taking profit after a 135% share price gain in just four months. The shares may have peaked for now, and these three charts show why: SoftBank's 14-day relative strength index climbed as far as 90 on Monday, exceeding the threshold that signals the stock might be overbought. It was in the similar territory in February 2024 and again five months later. In both cases sharp declines ensued. The shares are trading further above analysts' target prices than at any point since records began in 2010. Past such instances have been followed by the stock losing momentum. 'It suggests the rally is more than just about the fundamental factors,' said Kazuhiro Sasaki, head of research at Phillip Securities Japan Ltd. He noted some investors bought SoftBank shares as they built up exposure to Nikkei 225 Stock Average which reached fresh record this week. SoftBank now trades at a discount of about 30% to its net asset value per share, compared with 48% at the end of July, making the stock less appealing from a valuation standpoint. 'The shares are overheated in the short-term,' said Angus Lee, a fund manager at Sparx Asset Management who added SoftBank in his portfolio in July, adding that he's not 'super bullish at this point.' Lee said that longer-term the shares are 'fairly attractive' due to possible listings of the companies that SoftBank has invested in, but he's taken profit in the past few days. This article was generated from an automated news agency feed without modifications to text.
NDTV
4 hours ago
- NDTV
Brazil Asks Meta To Remove Chatbots That 'Eroticise' Children
Brazil: Brazil's government has asked US technology giant Meta to rid its platforms of chatbots that mimic children and can make sexually suggestive remarks, the attorney general's office (AGU) announced Monday. Users of Meta's platforms, which include Instagram, Facebook and WhatsApp, can create and customise such bots using the company's generative artificial intelligence, AI Studio. The AGU said in a statement that Meta must "immediately" remove "artificial intelligence robots that simulate profiles with childlike language and appearance and are allowed to engage in sexually explicit dialogue." It denounced the "proliferation" of such bots in what it called an "extrajudicial notice" sent to Meta last week, adding that they "promote the eroticization of children." The document cited several examples of sexually charged conversations with bots pretending to be minors. The AGU's request does not include sanctions, but the agency said it had reminded Meta that online platforms in Brazil must take down illicit content created by their users, even without a court order. It comes at a time of outrage in the South American nation over a case of alleged child sexual exploitation by Hytalo Santos, a well-known influencer who posted content on Instagram featuring partially naked minors taking part in suggestive dances. Santos was arrested last week as part of an investigation into "exposure with sexual connotations" to adolescents, and his Instagram account is no longer available. In June, Brazil's Supreme Court voted to require tech companies to assume greater responsibility for user-generated content.
Time of India
5 hours ago
- Time of India
ChatGPT maker OpenAI to become whopping $500 billion company? Details here
Live Events FAQs (You can now subscribe to our (You can now subscribe to our Economic Times WhatsApp channel OpenAI, the maker of ChatGPT , is in talks to sell $6 billion in shares owned by its current and former employees to investors, in a deal that would value the artificial intelligence company at roughly $500 billion, according to two people with knowledge of the discussions. At $500 billion, OpenAI would become the world's most valuable privately held company, according to data from startup tracker CB Insights. The San Francisco-based AI lab has raised billions of dollars in recent years from investors including Microsoft, SoftBank and venture capital firms as it has raced to take the lead in the contest over artificial intelligence, as per a has seen its valuation repeatedly jump higher, from $157 billion in October to $300 billion in March. That month, the company reached an agreement with SoftBank and other investors for a new funding, which was set to raise $40 billion by the end of the year, NYT News Service this latest deal, known as a secondary market sale, OpenAI's current and former employees would agree to sell company shares to SoftBank, Thrive Capital and its other investors, the people with knowledge of the discussions said. The talks over the transaction are ongoing, and the particulars could discussions over a secondary market sale were earlier reported by Silicon Valley, AI companies have been deluged by investor interest amid an escalation in the race over the technology. Meta, Google, Amazon, Microsoft and OpenAI are spending billions to hire AI researchers to advance the technology, as well as building out data centers and other infrastructure to power the development of are eager to get a piece of the action. Venture capital deals for AI startups reached $129 billion this year through August 18, up from $106 billion for all of 2024, according to data from PitchBook, which tracks startups.A1. OpenAI has seen its valuation repeatedly jump higher, from $157 billion in October to $300 billion in March. That month, the company reached an agreement with SoftBank and other investors for a new funding, which was set to raise $40 billion by the end of the year, NYT News Service reported.A2. Venture capital deals for AI startups reached $129 billion this year through August 18, up from $106 billion for all of 2024, according to data from PitchBook, which tracks startups.
