Meta's Llama 3.1 AI model reproduces copyrighted book text, raising legal alarms
Meta's latest AI model, Llama 3.1, is under scrutiny after a new study revealed it can replicate large portions of copyrighted books—including Harry Potter and the Sorcerer's Stone—with surprising accuracy. Conducted by researchers from Stanford, Cornell, and West Virginia University, the study found that Llama 3.1 has memorized around 42% of the first Harry Potter book and can reproduce 50-word sections correctly nearly half the time.
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
Time of India
3 hours ago
- Time of India
Iran urges public to delete WhatsApp; claims app sends data to Israel amid conflict; company denies allegations
Iranian state television on Tuesday urged its citizens to remove WhatsApp, alleging that the messaging app collects user data and shares it with Israel, amid the ongoing conflict, without providing any specific evidence. Tired of too many ads? go ad free now In response, WhatsApp strongly denied the accusations, saying, "We are concerned these false reports will be an excuse for our services to be blocked at a time when people need them the most." The Meta-owned app reiterated that it uses end-to-end encryption and does not share user data. "We do not track your precise location, we don't keep logs of who everyone is messaging and we do not track the personal messages people are sending one another," the company further clarified. "We do not provide bulk information to any government," it added. End-to-end encryption ensures that messages remain unreadable to anyone except the sender and intended recipient. Any intercepted messages appear as unintelligible text that cannot be decoded without the proper key. Meta Platforms, which also owns Facebook and Instagram, is WhatsApp's parent company. Throughout recent years, Iran has restricted access to various social media platforms, though many residents circumvent these restrictions using proxies and VPNs. The government blocked WhatsApp and Google Play in 2022 during widespread demonstrations following the death of a woman in morality police custody. These restrictions were lifted in late 2023. WhatsApp remains one of Iran's most widely used messaging applications, alongside Instagram and Telegram.
Mint
3 hours ago
- Mint
Musk's xAI Burns Through $1 Billion a Month as Costs Pile Up
Elon Musk's artificial intelligence startup xAI is burning through $1 billion a month as the cost of building its advanced AI models races ahead of the limited revenues, according to people briefed on the company's financials. The rate at which the company is bleeding cash provides a stark illustration of the unprecedented financial demands of the artificial intelligence industry, particularly at xAI, where revenues have been slow to materialize. To cover the gap, Musk's startup is currently trying to raise $9.3 billion in debt and equity, according to people briefed on the deal terms, who asked not to be identified because the information is private. But even before the money is in the bank, the company has plans to spend more than half of it in just the next three months, the people said. Over the course of 2025, xAI, which is responsible for the AI-powered chatbot Grok, expects to burn through about $13 billion, as reflected in the company's levered cash flow, according to details shared with investors. As a result, its prolific fundraising efforts are just barely keeping pace with expenses, the people added. A spokesperson for the company declined to comment. The losses are due, in part, to the huge costs that all AI companies face as they build the server farms and buy the specialized computer chips that are needed to train advanced AI models like Grok and ChatGPT. Carlyle Group Inc. estimates that over $1.8 trillion of capital will be deployed by 2030 to meet that demand to build out AI infrastructure, Chief Executive Officer Harvey Schwartz wrote in a shareholder letter. 'Model builders will look to raise debt and they're going to burn lots and lots of cash,' said Jordan Chalfin, senior analyst and head of technology at CreditSights. 'The space is very competitive and they are battling for technical supremacy.' But Musk's entrant in the AI race has struggled to develop revenue streams at the same rate as some of its direct competitors, like OpenAI and Anthropic. While almost none of these companies offer public figures on their finances, Bloomberg has previously reported that OpenAI, the creator of ChatGPT, is expecting to bring in revenues of $12.7 billion this year. At xAI, revenues are expected to be just $500 million this year, rising to north of $2 billion next year, investors were recently told. What xAI has on its side is a CEO, Musk, who is the richest man in the world, and one who has shown a willingness to spend his fortune on huge, futuristic projects long before they start generating money. Back in 2017, Musk's biggest company, Tesla Inc. was burning through $1 billion a quarter to pay for the production of its Model 3 car, Bloomberg reported at the time. SpaceX, meanwhile, sustained years of steady losses as it pushed toward its long-term goal of interplanetary exploration. Even against this backdrop, though, the huge losses at xAI stand out. Musk's team at xAI, which is racing to develop AI that can compete with humans, believes that they have advantages that will eventually allow them to catch up with peers. While some competitors rent chips and server space, xAI is paying for much of the infrastructure itself and getting direct access through Musk's social media company, X, which previously bought a significant stockpile of the most coveted and high powered computer chips. Musk has said that he expects xAI will continue buying more chips. After recently merging with X, Musk's AI executives are also hopeful that they will be able to train the company's models on the social media network's copious and constantly-refreshed archives, rather than paying for data sets like other AI companies. These potential advantages have led xAI to optimistically project that it will be profitable by 2027, people familiar with the matter said. OpenAI expects to be cashflow positive by 2029, Bloomberg previously reported. These projections, along with Musk's celebrity status and political power, have been enough to win over many investors, especially before the recent breakdown in the relationship between Musk and President Donald Trump. Potential xAI investors were told that the company's valuation grew to $80 billion at the end of the first quarter, up from $51 billion at the end of 2024. Investors have included Andreessen Horowitz, Sequoia Capital and VY Capital. For now, though, xAI is racing to raise enough money to keep up with its prodigious expenditures. Between its founding in 2023 and June of this year, xAI raised $14 billion of equity, people briefed on the financials said. Of that, just $4 billion was left at the beginning of the first quarter, and the company expected to spend almost all of that remaining money in the second quarter, the people said. The company is now finalizing $4.3 billion in new equity funding, and it already has plans to raise another $6.4 billion of capital next year, the company has told investors. And that is on top of the $5 billion in debt that Bloomberg has previously reported Morgan Stanley is helping it raise. The corporate debt is expected to help pay for xAI's data center development, the people said. Other companies have decided to do project financing instead. The company is also expecting to get a bit of help from a $650 million rebate from one of its manufacturers, it told investors this week. There were early signs that investors were hesitant to loan the company money at the proposed terms, Bloomberg has reported. The company gave select investors more detailed financial information on Monday in response to questions it had faced during the fundraising process, people familiar with the negotiations said. But the deal has attracted more interest since the company changed some of the deal terms — to be more investor friendly — and finalized the equity fundraising. A spokesperson for Morgan Stanley, the bank in charge of xAI's debt sale, declined to comment. With assistance from Tom Contiliano and Peter Pae. This article was generated from an automated news agency feed without modifications to text.
&w=3840&q=100)
Business Standard
4 hours ago
- Business Standard
OpenAI brings image generation feature for all through ChatGPT on WhatsApp
OpenAI will make the image generation feature available for WhatsApp users as well, said the company. 'This is the same image generation tool available in the ChatGPT web and mobile apps, now integrated into WhatsApp. Image generation in WhatsApp is free to use and is now available everywhere ChatGPT in WhatsApp is supported,' the company said in a statement. The integration of ChatGPT on WhatsApp was announced by OpenAI in December 2024, where users could send a message to ChatGPT to get up-to-date answers. Earlier this year, in April, ChatGPT said it would start allowing users to search, compare, and buy products in ChatGPT by giving users personalised recommendations for products, visual details of the product they are looking for, the price, as well as a direct link to buy it. Before that, in February 2024, OpenAI said it was 'testing the ability for ChatGPT to remember things you discuss to make future chats more helpful.' The idea, OpenAI had then said, was to save users 'from having to repeat information' so that future conversations with the chatbot become more useful. The memory feature was rolled out to all users in September that year. In an update on April 10 this year, OpenAI said that memory in ChatGPT is now comprehensive, as the large language model (LLM) could — in addition to memories saved by users — also reference past conversations between the chatbot and the user to deliver more personalised responses.
