AI’s hunger for more data, fueled largely by pirated content owned by others, has been recognised as being “fair use” by a court in the US.
However, this first major case on copyright and generative artificial intelligence also found that not paying for books to feed the AI is theft.
The owners of content, including major media players such as News Corp (which has already done deals with AI companies), say they want to be compensated for helping AI companies become multi-billionaires.
A court has found that AI company Anthropic is okay to train its machine brain, Claude, on copyrighted works as this falls into a definition under US law of “fair use” because the original content is transformed.
“The training use was a fair use,” senior district judge William Alsup said in his judgement on the class action brought by book authors.
“The technology at issue was among the most transformative many of us will see in our lifetimes.”
However, the judge said the artificial intelligence firm downloaded, without payment, millions of copyrighted books in digital form from pirate sites on the internet.
And it is on this point, the judge said, the authors could take the AI company to court. And the compensation for that, according to multiple reports, could run into the billions of dollars with each infringement attracting a penalty of up to $US150,000.
The court was told Anthropic cofounder, Ben Mann, downloaded at least five million copies of books from Library Genesis, or LibGen, which he knew had been pirated. At least two million copies of books came from the Pirate Library Mirror, also which Anthropic knew had been pirated
Anthropic also spent millions of dollars buying millions of print books which were stripped from their bindings and scanned the books into digital form.
“That Anthropic later bought a copy of a book it earlier stole off the internet will not absolve it of liability for the theft but it may affect the extent of statutory damages,” the judge said.
US copyright law says that wilful copyright infringement can result in damages of up to $150,000.
Under Australian copyright law, the concept of “fair use” doesn’t exist. Under section 42 of the Copyright Act, Australia has “fair dealing” which is an exemption for certain uses, such as reporting of news.
Josephine Johnston, the CEO of the collecting body Copyright Agency, says Australia has the advantage of existing statutory licences that are sufficiently broad and flexible to support AI development and use in the government and education sectors.
“They require fair payment to rights holders but are an efficient way to support use at large scale, providing easy and legal rights to use a vast amount of copyright material,” she said.
“Copyright Agency has also just launched an extension to its business licence that permits use of Australian news content in connection with AI tools to drive internal efficiencies and productivity – addressing key use cases such as summarisation and analysis.
“We are working with both rights holders and corporations to provide more licences to support AI business needs in Australia. Licensing is developing rapidly, both here and overseas.
Johnston said two cases illustrate how US doctrine of fair use fails to address the challenges of AI development, both in terms of real world impact on creators and providing certainty for developers.
“In the first, Anthropic supposedly emerged as the “victor” but not even fair use could save it from now facing millions of dollars in damages for the use of pirated books (a common practice in foundation model training where content is indiscriminately scraped and stolen),” she said.
“In the second case, Judge Chhabria was clear that his decision was confined to the specific facts before him and that fair use would often not succeed as a defence – an outcome that points to years of legal uncertainty and chaos as dozens of US cases are heard and appealed.
“Fortunately, fair use is not part of our legal framework.
“AI developers just need to ask – does anyone really believe they can’t afford it? Innovation and the productivity gains coming from it, can be supported by licences which provide certainty for technology companies and fair payment for the works used.
“The Australian government should resist the calls to water down our fair and flexible copyright system, favouring multinational tech companies over our creative sector.
“Those laws are essential to ensure that AI can flourish in Australia while everyone contributing essential inputs is properly compensated for their work.”
(DISCLOSURE: Chris Pash is a former board director of the Copyright Agency and of the Australian Society of Authors.)
Have something to say on this? Share your views in the comments section below. Or if you have a news story or tip-off, drop us a line at adnews@yaffa.com.au
Sign up to the AdNews newsletter, like us on Facebook or follow us on Twitter for breaking stories and campaigns throughout the day.