Anthropic Australia: OpenAI Seek Eased Copyright Rules for AI Training

The Anthropic Australia dispute now covers a copyright framework created in 1968, as Anthropic and OpenAI submitted joint recommendations to Australian officials in Canberra during September 2026. The two San Francisco-based AI companies want the government to reconsider restrictions on using local copyrighted material to train artificial intelligence systems.

Australia’s Attorney-General’s Department had previously launched a consultation paper on copyright and artificial intelligence in the second quarter of 2023. That earlier process established the policy question now facing lawmakers: how can Australia protect writers, musicians, publishers, filmmakers, and other rights holders without cutting local material out of AI development?

Why the Anthropic Australia copyright dispute matters

The rules sit within the Australian Copyright Act 1968, which governs how protected creative works can be copied and used. Australia’s current approach requires AI developers to obtain permission from copyright owners before using those works for training, according to the submissions described by Thenextweb. That requirement creates a difficult open.

What Anthropic and OpenAI asked Australia to change

Anthropic accepted that Australia had ruled out a broad copyright exception for AI training, but proposed a narrower system based on conditional approval. Those conditions could include investment or other support for Australian creators and cultural production, rather than unrestricted access to copyrighted material.

OpenAI called for a balanced framework that would allow models to learn from publicly available information while giving rights holders practical opportunities to work with AI companies. That proposal leaves open several questions, including whether “publicly available” should mean merely accessible online or clearly licensed for machine learning.

The distinction matters. Anthropic’s approach places more emphasis on conditions attached to permission, while OpenAI’s proposal focuses on a framework that combines access with collaboration. Neither submission removes the need for rights holders to understand how their work enters training datasets.

How Australia compares with other AI copyright systems

Australia’s framework is reportedly stricter than approaches used in the United Kingdom and Japan, although the supplied comparison remains unconfirmed. That difference could affect where AI companies build partnerships, conduct data work, and develop products that need locally relevant language or cultural material. The issue also connects to wider legal pressure on the sector.

Our coverage of the Lawsuit Anthropic, OpenAI, dispute shows why training practices now attract scrutiny beyond technical performance. A separate report on Anthropic Reportedly GPT-6 illustrates the commercial value attached to increasingly capable models, but capability does not settle ownership questions. Creators have a legitimate concern: if a model learns from their work, a general opt-out process may not provide meaningful control. AI companies have an equally specific concern: a permission system covering every individual work could make large-scale training slow, expensive, or impossible.

What happens next for local AI training?

The Department of Infrastructure, Transport, Regional Development, Communications and the Arts will need to assess whether conditional access can be enforced without creating a loophole. A workable framework would need transparent records, clear rights-holder contact methods, and consequences when companies use protected material outside agreed terms. OpenAI’s public policy materials provide useful context on how the company presents its approach to regulation, while Anthropic’s proposals add pressure for a model that recognises cultural investment as part of access.

The anthropic Australia debate will therefore move beyond the two companies’ submissions. Australia’s eventual decision could influence negotiations between AI developers and creators in other countries, especially where governments are writing copyright rules after generative AI has already become widely used. The next Australian framework should make training permission traceable, compensation testable, and public-interest access narrow enough to protect creators.

FAQs

What is the Anthropic Australia copyright dispute?

It is a policy dispute over whether Anthropic and OpenAI should receive limited access to Australian copyrighted content for AI training under the Australian Copyright Act 1968.

Where did the companies submit their recommendations?

Representatives submitted joint policy recommendations to the Department of Infrastructure, Transport, Regional Development, Communications and the Arts in Canberra.

What did Anthropic propose?

Anthropic proposed a narrow form of conditional approval that could include investment or support for Australian creators and cultural work.

What did OpenAI request?

OpenAI requested a balanced framework allowing models to learn from publicly available information while giving rights holders opportunities to cooperate with AI companies.