Anthropic and OpenAI call on Australia to ease ban on learning AI models

Two leading developers in the field of artificial intelligence (AI) OpenAI and Anthropic appealed to the authorities of Australia to reconsider the ban on the use of creative content of the country to train their models.

The day before, Canberra announced that it would put in place a broad set of rules regarding AI and data centers in 2027. These include a clause in copyright law that effectively prevents mostly US-based companies from training on Australian material there.

Companies are asking for exemptions to protect the rights of content creators while encouraging investment. Anthropic is proposing that the government allow AI models to be trained “through narrow formal approval” and stipulates its “investment or other conditions to support the future of Australian authors and cultural endeavors.” OpenAI proposes a “balanced copyright system that allows models to learn from publicly available information.”.

OpenAI has previously signed an agreement with Australian data center developer NextDC for a purchase for a planned data center in Sydney. And Anthropic last week was unveiled as a partner for such a centre in Queensland.

The Australian government is actively pitting the interests of AI developers against the interests of authors, publishers and other content creators who have opposed any exemptions in the bill. By exception, creators understand loopholes to use their works to train commercial AI models without their consent or payment.

The Australian Parliament’s Joint Select Committee on AI, set up in August, should present its vision of the situation. A report by lawmakers is expected in November.

Ernest Philippowski