Reddit CEO Steve Huffman on the floor of the New York Stock Exchange during Reddit’s initial public offering on March 21, 2024 in New York City. (Photo: Spencer Platt/Getty Images)Get ready to rumble.
Reddit sued Anthropic on Wednesday, alleging that the AI startup was illegally “scraping” Reddit comments to train its chatbot.
The popular social media site alleges that Anthropic used automated bots to access the content despite being asked not to do so, and failed to get users’ consent.
“AI companies should not be allowed to scrape information and content from people without clear limitations on how they can use that data,” Reddit’s chief legal officer said.
Anthropic rejected the claims. “We disagree with Reddit's claims and will defend ourselves vigorously,” it said in a statement.
Reddit has licensing agreements with Google, OpenAI, and other Anthropic rivals to receive compensation in exchange for the companies’ commercial use of the data—that is, training their AI systems on Reddit comments.
But not Anthropic.
The lawsuit cites a 2021 paper co-authored by Anthropic CEO Dario Amodei that describes Reddit’s user forums, or “subreddits,” as a source of high-quality AI training data.
A 2023 Anthropic letter to the U.S. Copyright Office argued that the way its chatbot, Claude, trains “qualifies as a quintessentially lawful use of materials” because it copies information to perform statistical analysis.
But Reddit’s lawsuit doesn’t allege copyright infringement, just a breach of its terms of use.
—AN