Topline
Reddit on Wednesday sued Anthropic for alleged breach of contract, claiming the AI startup trained its models on Reddit users’ personal data without permission and continued to do so despite telling Reddit, which has struck licensing deals with OpenAI and Google, the company had stopped.
Anthropic accessed Reddit’s servers more than 100,000 times after saying it stopped, Reddit alleges. … More
Key Facts
Reddit’s lawsuit, filed Wednesday in the Superior Court of California, County of San Francisco, accused Anthropic of “unlawful and unfair business acts” after the company “intentionally trained” its models on Reddit user data “without ever requesting their consent” or paying for it.
Reddit warned Anthropic against using its data and attempted to engage in licensing negotiations, though Anthropic refused and told Reddit the company would stop accessing the social media platform’s servers, Reddit claims.
Anthropic continued to access Reddit’s servers more than 100,000 times after the company told Reddit otherwise, according to the complaint.
Reddit requested damages to be determined after a jury trial, and an injunction that would block Anthropic from accessing Reddit user data or using older data for new products.
Anthropic did not immediately respond to a request for comment.
Crucial Quote
“Anthropic is in fact intentionally trained on the personal data of Reddit users without ever requesting their consent,” Reddit claims, calling Anthropic a “late-blooming” firm that “bills itself as the white knight of the AI industry.”
Tangnet
Reddit’s stock jumped more than 7% to nearly $119 by Wednesday afternoon. Shares of Reddit, whose initial public offering was held in March 2024, have dropped 28% on the year despite peaking at an all-time high of more than $225 in February.
Key Background
Reddit’s decades-old forums have been the target of several AI companies for training models in recent years. CEO and cofounder Steve Huffman told The Wall Street Journal last year that Reddit initially gave away its data for free before realizing AI firms were engaged in an “arms race,” later saying Reddit was “in talks with just about everybody” for licensing deals. AI firms rely on “actual intelligence” to train their models, Huffman said, “which is what you find on Reddit” across hundreds of thousands of subreddit forums. OpenAI and Reddit announced a partnership in May, which OpenAI said would “enhance” its AI chatbot ChatGPT. The deal followed a $60 million agreement for Google to train its AI models using Reddit’s posts, and Reddit has reached data access partnerships with software firms Sprinklr and Cision.
Further Reading
