Artificial Intelligence
Advertisement
Supported by
In its lawsuit, Reddit said Anthropic had also declined to enter into a licensing agreement for data and had unjustly enriched itself at Redditâs expense.
By Mike Isaac
Reporting from San Francisco
Reddit sued Anthropic on Wednesday, accusing the artificial intelligence start-up of unlawfully using the data of Redditâs more than 100 million daily people to train its A.I. systems.
The lawsuit, which was filed in the Superior Court of California in San Francisco, claimed that Anthropic had obtained access or tried to obtain access to Reddit data more than 100,000 times, in a breach of the online platformâs content policies. Anthropic also declined to enter into a licensing agreement for the data, the lawsuit said, and unjustly enriched itself at Redditâs expense.
âWe will not tolerate profit-seeking entities like Anthropic commercially exploiting Reddit content for billions of dollars without any return for redditors or respect for their privacy,â Ben Lee, Redditâs chief legal officer, said in a statement. âA.I. companies should not be allowed to scrape information and content from people without clear limitations on how they can use that data.â
A spokeswoman for Anthropic did not immediately provide a comment.
The lawsuit was the latest clash over the use of digital data by A.I. companies amid a heated race to develop the Tech. For years, A.I. companies have gobbled up as much data across the internet as possible to fine-tune their systems, which rely on information to improve the responses they generate. But those data sources are quickly drying up, as companies lock down more of their data to keep it from being used without permission.
Reddit, which is 20 years old and went public last year, is among a generation of social media companies with a wealth of user-generated conversational data. The site is a kind of social message board, where people can hold discussions on any number of topics, from dogs to television shows to cryptocurrency.
In recent years, Reddit executives began realizing just how valuable the companyâs data was to the rest of the industry. Steve Huffman, Redditâs chief executive, began talking to companies like Google and OpenAI to potentially strike licensing deals. Reddit eventually reached deals with Google and OpenAI for access to Reddit public conversation data to train their A.I. systems for a fee.
Anthropic, a high-profile A.I. start-up in San Francisco, makes the Claude chatbot and is regarded as a leader in the industry. In March, it completed a new fund-raising deal that valued it at $61.5 billion, up from about $16 billion a little more than a year ago. Dario Amodei, Anthropicâs chief executive, and his sister, Daniela Amodei, its president, founded the company as a start-up that would build A.I. with guardrails.
Anthropic has avoided negotiating with Reddit to license the online platformâs data, according to Redditâs lawsuit. The start-up instead has continued accessing â known as scraping â Redditâs public conversational data, according to the suit.
âA.I. companies should not be allowed to scrape information and content from people without clear limitations on how they can use that data,â Mr. Lee said in a statement. âLicensing agreements enable us to enforce meaningful protections for our people, including the right to delete your content, user privacy protections and preventing people from being spammed using this content.â
In 2023, The New York Times sued OpenAI and its partner, Microsoft, for copyright infringement, accusing the tech companies of using millions of its news articles without authorization. Last week, The Times said it had struck a deal with Amazon to license its editorial content for use in the tech giantâs A.I. platforms.
Mike Isaac is a Tech correspondent for The Times based in San Francisco. He regularly covers Facebook and Silicon Valley.
News and Analysis
Investing in Data Centers: Private equity firms like Blackstone are using their clientsâ money to buy and build data centers to fuel the artificial intelligence boom.
Energy Dept. Supercomputer: The flagship supercomputer, planned for 2026, will use Nvidia chips tailored for A.I. calculations and the simulations common to energy research and other scientific fields.
The Times and Amazon Deal: In 2023, The Times sued OpenAI and Microsoft for copyright infringement. Now its editorial content will appear across Amazon platforms.
Nvidia Earnings: Sales in its most recent quarter rose 69 percent from a year earlier to $44.1 billion. Its net income rose 26 percent to $18.78 billion.
The Age of A.I.
Tool for Chefs:Â Some restaurateurs are starting to explore ways artificial intelligence can help them create recipes, menus and dining experiences.
Google's New Search: AI Mode excels at tasks like product research for online shopping. But it falls short on basic web searches.
Reading List in Chicago Sun-Times: An A.I.-generated summer reading insert recommended made-up titles by real authors such as Isabel Allende and Delia Owens.
Advertisement