Not a single multinational AI platform operating in Australia has approached a major music licensing body to negotiate training data agreements in four years, even as these companies invest billions in data centers and claim they want to operate responsibly. This striking absence of licensing requests forms the centerpiece of a formal submission to Australia's Joint Select Committee on Artificial Intelligence, raising urgent questions about whether the world's most powerful tech companies are deliberately avoiding copyright obligations.

Why Are AI Companies Refusing to License Creative Content?

APRA AMCOS, which represents over 128,000 songwriters, composers, and music publishers and distributed almost $700 million in royalties last year, submitted formal testimony asking lawmakers to demand a direct answer from every AI platform and tech industry body appearing before the committee: Why are they refusing to enter into licensing arrangements for high-value creative content ?

The organization notes that licensing is not a technical barrier. As of January 2026, at least 274 known commercial AI licensing agreements exist worldwide covering music, news, books, images, and video, all negotiated under copyright laws no more permissive than Australia's. These deals prove that licensing AI training on copyrighted material is commercially viable and legally feasible.

"Multinational AI platforms are the most powerful companies the world has ever seen, and they are throwing that weight around to get the rules of the global IP market rewritten in their favour. Australia is where they've decided to start," said Dean Ormston, CEO of APRA AMCOS.

Dean Ormston, CEO of APRA AMCOS

According to APRA AMCOS, the platforms have invested heavily in lobbying and infrastructure but are conspicuously absent from licensing negotiations. "They have found the money for nearly every lobbying firm in Canberra. They have found the money for data centres, power and Nvidia chips. But they have no intention of paying for the single most important ingredient in training and running a frontier large language model, which is creative content," Ormston stated.

What Do Internal Microsoft Documents Reveal About AI Training Practices?

The stakes became clearer when court filings were unsealed in The New York Times' lawsuit against OpenAI and Microsoft. A senior Microsoft executive described the companies' AI training practices as "an astonishing theft of unprecedented proportions," calling it perhaps the "largest theft of labor in human history." OpenAI's own leadership privately described its models as an "existential threat" to the creators whose work trained them.

These admissions from inside the companies themselves contradict public claims about responsible AI development. APRA AMCOS emphasized the contradiction: "We have called this theft for four years. Now we know Microsoft's own executives were calling it the same thing in private".

More than 140 copyright lawsuits have been filed against AI companies in the United States, making Australia an increasingly attractive destination for platforms seeking regulatory environments more favorable to their interests. One major platform has reportedly tied a $21.6 billion Australian data center investment to copyright questions being settled in its favor.

How to Protect Creative IP in the AI Era: Key Policy Recommendations

APRA AMCOS has proposed seven specific recommendations to the Australian government to ensure creative industries are not undervalued in the rush to attract AI investment:

• Training Data Disclosure: Require any AI model made available in Australia to disclose its training data sources, allowing regulators and rights holders to verify compliance with copyright law.
• New Legal Cause of Action: Create a standalone legal remedy against models built on infringing uses, regardless of where the training took place, closing loopholes for offshore training operations.
• Indigenous IP Legislation: Establish dedicated Indigenous Cultural and Intellectual Property legislation with digital and AI provisions brought forward as a priority.
• Sovereign AI Fund: Establish a $500 million Commonwealth fund for investment in lawful, sovereign AI development specifically in the creative sector, ensuring Australia builds AI capacity without compromising creator rights.

The submission reflects a century of experience managing technology licensing. APRA AMCOS was established six months after the first radio broadcast in Australia by music publishers who understood that new technologies would use creative work. Over 100 years, the organization has successfully licensed every major technology from radio and cassette tape to downloads, streaming, and social media platforms.

"We've been doing this for a hundred years, and we've watched plenty of technology companies arrive convinced the rules didn't apply to them," Ormston noted. "Radio worked it out. Television worked it out. The download stores, streaming services and social media platforms all worked it out and built businesses worth billions based on licences".

What Is the Global Context for AI Copyright Disputes?

Australia's copyright standoff with AI companies is not isolated. China's Supreme People's Court issued its first comprehensive judicial guidance on AI disputes in September 2026, establishing a framework that allocates copyright liability among developers, service providers, and users based on control, participation, and precautions taken.

Under China's framework, when AI-generated content infringes copyright, developers invoking a non-infringement defense may be required to disclose training data sources, training records, model operation modes, and supporting technical evidence. This evidentiary approach acknowledges that rights holders cannot inspect a model's internal workings but stops short of presumptively holding every developer liable.

The Chinese guidance deliberately leaves two critical questions unresolved: whether AI-generated content itself can qualify for copyright protection, and whether training models on copyrighted works without consent constitutes infringement. The court concluded that further case-by-case experience was needed before committing to hard rules on these foundational issues.

Meanwhile, the U.S. Patent and Trademark Office (USPTO) appointed Jonathan Spencer as its first Chief Artificial Intelligence Officer in September 2026, signaling heightened focus on how AI intersects with patent and intellectual property systems. Spencer brings experience from Waymo building safety-critical AI models and from multiple startups applying AI to healthcare and telecommunications.

As governments worldwide grapple with AI copyright frameworks, Australia's decision will likely set a precedent. APRA AMCOS warns that the creative industries contribute economically, socially, and culturally to Australia, and must not be "undervalued and sold off expeditiously" in pursuit of AI investment.