Writers sue Anthropic for ‘stealing’ their work for AI training
Joining the ranks of Microsoft and OpenAI, Anthropic is now also facing a lawsuit from authors who claim that the creator of the Claude AI models used the writers’ work to train its LLMs unsolicited and, more importantly, without paying for it.
The trio of writers filed the suit in a court in the U.S. state of California. According to them, Anthropic abuses ‘human creativity and ingenuity’ by feeding the fruits of their labor (and that of others) without permission to the AI models it trains.
“Humans who learn from books buy lawful copies of them, or borrow them from libraries that buy them, providing at least some measure of compensation to authors and creators,” the complaint reads. Anthropic has not responded apart from saying it is ‘aware’ of the suit, Reuters reports.
Motley crew of plaintiffs
The authors who sued Anthropic are Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson. All three write about very different things: thriller and mystery stories in Bartz’s case, non-fiction in Graeber’s, and Johnson’s published about the war in Iraq after 9/11 and then turned to a true-crime report about the theft of exotic bird feathers for the purpose of fly fishing(!)
Not only have book writers set their sights on Antropic. Music publishers have also sued the company because its AI models would regurgitate existing song lyrics based on users’ prompts. In other words, these have allegedly also been used for data entry.
Rival to OpenaI
Founded in 2021, Anthropic positions itself as a rival to OpenAI. Its models in the Claude series, including Sonnet, Opus, and Haiku, are said to be able to compete with the various models of its well-known San Francisco neighbour, including GPT-4o, GPT-3.5 Turbo, and -4o mini.
Sam Altman, the much-discussed founder of OpenAI and just about the only one of the original team of founders still left, claims that AI models cannot be properly trained without copyrighted data. That means books, newspaper articles, scientific papers and other data sources containing high-quality information.