Statement against AI training passes 50,000 signatures

More than 50,000 creators from the worlds of music, film and television, literature and academia have signed on to a statement condemning the use of creative work to train generative AI models.

Actors Charles Dance and Julianne Moore, writers Emily St John Mandel and Jeffrey Archer, and musicians Thom Yorke and Roger Daltrey are among the signatories to the statement, which warns the use of their work to train AI models presents a “major, unjust threat” to their livelihoods.

The statement originally launched with 10,500 signatories in October last year:

“The unlicensed use of creative works for training generative AI is a major, unjust threat to the livelihoods of the people behind those works, and must not be permitted.”

Tensions in AI training

It comes amid a flurry of lawsuits lobbed by creative professionals at AI firms, calling out the use of their work without permission to develop AI products.

In January, court documents revealed that Facebook parent company Meta used a pirated database of more than 7 million books and 81 million academic papers to train its Meta AI model.

At no point did Meta seek permission to use the authors’ works — which include books by Australian authors including Charlotte Wood, Alexis Wright, Tim Winton and Helen Garner — to train its AI systems, or offer compensation.

It wasn’t the first time either, with pirated works by novelists including Winton, Peter Carey, and Jane Harper used to train multiple generative AI models in 2023.

Australian author Jennifer Mills sits on the board of the Australian Society of Authors (ASA), one of almost 300 organisations that have signed onto the statement.

She said in addition to the threat to artists’ livelihoods, there was also an emotional aspect at play.

“You spend years perfecting a book, working so hard, putting your heart and soul into something, and then to see it on a list of something that’s been stolen is just really heartbreaking,” Ms Mills said.

Industry stalwart OpenAI has also come under fire in the past week after launching a feature in their ChatGPT model which converts photographs to Studio Ghibli-style “art” — an artstyle pioneered by Hayao Miyazaki, whose opinions on AI-generated creations are well-known.

‘Theft’: What’s Australia doing?

Federal Arts Minister Tony Burke told ABC Arts last week that there was nothing the Australian government could currently do to prevent unlicensed use of creatives’ work to train AI models, a practice Mr Burke viewed as “theft”.

“The nature of this intellectual property is that once AI has used it, there’s no concept of return. It’s done,” he said.

He said regulation and protection for artists has lagged behind technological innovation and addressing the problem needed to be a collaborative global affair.

“So we need the international system to catch up pretty quickly with where technology is taking us,” he said.

“There’s all the difference in the world between an artist … using it as a tool of their art, and someone who thinks that they’ve just written something for a publisher and suddenly discovering that some other company’s just taken it.”

Ms Mills said she didn’t believe that the Australian government was powerless to intervene.

She said the ASA wanted to see the government legislate to protect creators and force tech companies to act in good faith and “do the right thing by [authors]”.

“They can license our work for AI training, pay us a fee, and that could become something that generates an ongoing income for authors to offset the downward pressure that’s being put on our incomes and our livelihoods,” she said.