Anthropic and OpenAI ask Australia to ease its ban on AI training with local content

Anthropic and OpenAI have asked Australian lawmakers to reconsider rules that stop them from training AI models on the country’s creative work, Reuters reported on Tuesday, citing the companies’ submissions to a parliamentary inquiry into AI laws.

In October 2025, the Australian government stated that it would not include a text and data mining exception in its copyright legislation.

In effect, this means that AI developers must obtain permission from copyright holders to train their machines on those works, a situation that in practice prevents mostly US-based laboratories from training models in the country.

Both companies have acknowledged that decision in their submissions but are asking for some margin around it. Anthropic has acknowledged that broad copyright exceptions have been ruled out and has therefore suggested a ‘narrow form of conditional approval’ for the purpose of model training.

Also, it stated that it would be open to conditions such as investment or other measures intended for the support of Australian creators and cultural work; OpenAI requested a balanced framework which would enable the models to learn from publicly available information while at the same time providing rights holders with opportunities to work together with AI companies.

Both companies connected the question to their intended data center plans in Australia; OpenAI has a power purchase agreement with NextDC for a site it plans to build in Sydney, and Anthropic was announced last week as a partner in a data center project in Queensland.

Anthropic has established that connection previously; when briefing notes were made public as a result of a request under the Freedom of Information Act, it was stated that Anthropic’s chief executive, Dario Amodei, had raised the issue of copyright with Treasurer Jim Chalmers in April, according to AFP.

The officials noted that Anthropic would claim its investment in Australia depended on there being clarity regarding copyright settings.

The creative industries, on the contrary, have made a strong effort. In July, Sarah Hanson-Young, a Green senator and chairwoman of a Senate inquiry into AI and data centers, stated that AI companies wishing to train on copyrighted material “should pay for it just as everyone else does”.

Anthropic has also encountered this problem in court. In the United States, it agreed in 2025 to pay $1.5 billion to settle a class action brought by authors over books that had been pirated and used in its training data, following a ruling by a judge that training on lawfully acquired books could be considered fair use.

The submissions were handed to a joint select committee on AI which both houses of parliament established in August. At the same time, Australia intends to introduce more comprehensive AI and data center regulations next year.