The prospect of journalists, photographers and artists claiming copyright infringements against artificial intelligence has taken another hit with Getty Images dropping its case against Stability AI.
But there is a glimmer of hope from another US court where the data being used for training AI models is pirated and therefore a likely breach of copyright.
Stability AI operates Stable diffusion, an AI image making tool that is trained on pre-existing images. Getty argued that Stability AIs computer generated images was a breach of copyright and undermined the capacity of its professional photographers to make a living.
However, its claim appeared to have been thwarted from the start because it took its case to a UK court. American courts claim jurisdiction on AI copyright cases where the AI training takes part in the US which is most of the time.
US copyright law has proven to be weak in protecting works by a journalist, artist, photographers, and music composers.
“It was always anticipated to be challenging to prove that connection to the UK because we know that most of the training happened in the US,” AI legal expert Alex Shandro told Associated Press.
Yet there is a glimmer of hope for creatives generally in a judgment brought down in another case mounted by three authors, Andrea Bartz, Charles Graeber and Kirk Walker, who took action against Amazon-backed US startup Anthropic for training its AI chatbot Claude on their books.
The judge dismissed that particular claim, saying Anthropic was not obliged to seek the authors’ permissions, again citing ‘fair use’ under US law. The law dismisses claims of copyright infringements as fair use if its use was found to be transformative, which applied in this case.
“This order grants summary judgment for Anthropic that the training use was a fair use. And it grants that the print-to-digital format change was a fair use for a different reason,” San Franciso Judge William Alsup wrote.
But he took issue with Anthropic copying and storing more than seven million pirated copies of books, including books from the notorious Books3 database of 196,000 books. This database includes a large number of books by Australian authors.
Alsup has ordered a trial in December to examine and order compensation over the piracy breach which does attract a copyright infringement. Anthropic claims it stored the books but didn’t use them to train its AI models.
Nevertheless, if Anthropic is forced to pay up on this ground, it opens the door for Australian authors to prosecute big tech companies purely on piracy grounds. Here Individuals who commit piracy face fines of up to $117,500 per infringement while corporations pay up to $585,000. Whether such action takes place under US or Australian law remains to be seen.
Given weak US copyright law, it has been argued that countries outside the US would be better off passing laws that require AI developers to seek permission and offer payments for training their models on the works of others.
About Post Author
