Copyright: AI companies win two legal victories in the United States

In quick succession, American writers and publishers have just suffered two legal setbacks at the hands of artificial intelligence (AI) companies.

On Monday, June 23, a federal judge in San Francisco, California, ruled that Anthropic could legally use copyrighted works to train its AI model, Claude, “without permission from the authors,” the Los Angeles Times reported .

As the specialist site Tech Crunch analyses :

“This is the first time that a court has upheld AI companies' argument that using copyrighted content to train their large language models would constitute 'fair use ' .”

Then, on June 25, Meta won a similar victory. Writers, including Ta-Nehisi Coates , accused the parent company of Facebook and Instagram of using their works without permission to train its AI, LLaMa. Again, a US federal judge ruled that this was fair use, though the Financial Times makes an important point : in his ruling, Judge Vince Chhabria notes that he is less endorsing Meta's arguments than rejecting the prosecution's strategy. In other words: with better arguments, the plaintiffs could prevail.

It's unclear whether these initial decisions will be followed by other judicial institutions. But they do mark a warning shot, at first instance, for artists of all disciplines, according to Tech Crunch. "This sets a precedent; the courts could continue to rule in favor of tech companies over creators."