OpenAI, Anthropic Seek Narrow Exemption to Australia's AI Training Ban

OpenAI and Anthropic have asked Australia to soften its copyright rules so that their artificial intelligence models can be trained on the country's creative content, arguing that a limited exemption could safeguard artists' rights while drawing further investment.

The submissions were made to a parliamentary inquiry examining laws governing AI. The two companies acknowledged the existing ban, which effectively prevents them from carrying out model training in Australia, but pressed for it to be eased.

Anthropic, the developer of the Claude chatbot, said broad copyright exceptions had not drawn widespread support and had been ruled out by the government. As an alternative, it proposed that AI training be permitted through a narrow form of conditional approval. The company added that it was open to conditions requiring investment or other measures aimed at supporting Australian creators and cultural work.

OpenAI, which makes ChatGPT, called for a balanced copyright framework that would allow models to learn from publicly available information while giving rightsholders opportunities to collaborate.

Both firms linked the copyright question to large infrastructure projects they are pursuing in Australia. OpenAI has signed an offtake agreement with Australian data centre developer NextDC for a planned facility in Sydney, while Anthropic was named last week as a partner for a data centre in Queensland.

The submissions add to pressure on the government as it weighs the interests of AI developers against those of authors, publishers and other creators, who have opposed exemptions that would let their work be used to train commercial models without consent or payment.

Australia has said it will bring in a broad set of rules covering AI and data centres from next year. The Joint Select Committee on Artificial Intelligence is due to deliver its report in November.