An AI startup has agreed to a $2.2 billion copyright settlement. But will Australian writers benefit?

Anthropic, an AI startup founded in 2021, has reached a groundbreaking US$1.5 billion settlement (AU$2.28 billion) in a class-action copyright lawsuit. The case was initiated in 2024 by novelist Andrea Bartz and non-fiction writers Charles Graeber and Kirk Wallace Johnson.

If the settlement is approved by the presiding judge, the company will pay authors about US$3,000 for each of the estimated 500,000 books included in the agreement. It will destroy illegally downloaded books and refrain from using pirated books to train chatbots in the future.

This is the largest copyright settlement in US history, establishing a crucial legal precedent for the evolving relationship between AI companies and content creators.

It will have implications for numerous other copyright cases currently underway against AI companies OpenAI, Microsoft, Google, and most recently Apple. In June, Meta prevailed in a copyright case brought against it, though the ruling left open the possibility of other lawsuits.

The settlement follows a landmark US ruling on AI development and copyright law, issued in June 2025, that separated legal AI training from illegal acquisition of content. Anthropic allegedly pirated over seven million books from two online “shadow libraries” in June 2021 and July 2022.

The plaintiffs and Anthropic are expected to finalise a list of works to be compensated by September 15.

Cautious optimism

In Australia, the response to news of a potential settlement has been cautiously optimistic. Stuart Glover, head of policy at the Australian Publishers Association told me:

We welcome these court-enforced steps towards accountability, but this settlement shows why AI companies must respect copyright and pay creators – not just see what they can get away with.

And for the sake of Australian authors and publishers whose works have been unlawfully scraped without compensation under this action, it’s a clear call for the Australian government to maintain copyright and mandate that AI companies pay for what they use.

Lucy Hayward, CEO of the Australian Society of Authors, told me:

While all of the details are yet to be revealed, this settlement could represent a very welcome acknowledgement that AI companies cannot just steal authors’ and artists’ work to build their large language models.

However, in the Anthropic case, authors will be only compensated if their publishers have registered their work with the US copyright office within a certain timeframe. Hayward expressed concern about this, as the seven million works that are alleged to have been pirated were written by authors from around the world and “we suspect many international authors may miss out on settlement money.”

She has called on Australian the government to introduce new laws requiring tech companies to “pay ongoing compensation to creators where Australian books have been used to train models offshore”.

Legal risks

In June, US judge William Alsup ruled that using books to train AI was not a violation of US copyright law. But he ruled Anthropic would still have to stand trial over its use of pirated copies to build its library of material.

Judge Alsup has since criticised the settlement for its loopholes. He has scheduled another hearing for September 25. “We’ll see if I can hold my nose and approve it,” he said.

If the settlement is not approved, Anthropic risks significantly greater financial repercussions. The trial is scheduled for December. If the company loses the case, US copyright law allows for statutory damages of up to $150,000 per infringed work in cases of wilful copyright infringement.

William Long, a legal analyst at Wolters Kluwer, suggests potential damages in a trial could reach multiple billions of dollars, potentially jeopardising or even bankrupting the company.

Anthropic recently secured new funding worth US$13 billion, bringing its total value to $183 billion. Keith Kupferschmid, president and CEO of the US-based Copyright Alliance, has argued that this is evidence “AI companies can afford to compensate copyright owners for their works without it undermining their ability to continue to innovate and compete”.

For Mary Rasenberger, CEO of the Authors Guild, the historic settlement is “an excellent result for authors, publishers, and rightsholders”. Rasenberger expects “the settlement will lead to more licensing that gives author[s] both compensation and control over the use of their work by AI companies, as should be the case in a functioning free market society.”

A step forward

While this particular settlement may offer little help to Australian writers and publishers whose works are not registered with the US Copyright Office, overall it is at least potentially good news for creators globally. It represents a step towards the establishment of a legitimate licensing scheme.

Australian copyright law does not include a US-style “fair use” exception, which AI companies claim protects their training practices. There have been calls to change the law with major AI players, including Google and Microsoft, lobbying the Australian government for copyright exemptions.

The recent Productivity Commission interim report proposed a text and data mining exception to the Australian Copyright Act, which would allow AI training on copyrighted Australian work. The proposal faced strong opposition from the Australian Society of Authors and the publishing industry.

As Arts Minister Tony Burke stated in August 2025, the government has “no plans, no intention, no appetite to be weakening” our copyright laws.

The Australian publishing industry is not entirely opposed to AI, but significant legal and ethical challenges remain. The Australian Publishers Association has advocated for government policies on AI that prioritise a clear ethical framework, transparency, appropriate incentives and protections for creators, and a balanced policy approach, so that “both AI development and cultural industries can flourish”.

Agata Mrva-Montoya is President of the Round Table on Information Access for People with Print Disabilities