The Office of AI is a template, not a plan
On 15 July 2026, the prime minister, Anthony Albanese, delivered the speech, AI in Australia’s interests. In this 30-minute address from The University of Sydney, home to the AI, Trust and Governance Centre, Albanese proposed that Australia would become the ‘first country in the world’ to have a single national framework of AI governance. While the speech recognised some of Australians’ concerns about commercial AI, the government’s proposed framework falls short of a comprehensive, robust, ethical response to ‘decid[ing] what AI looks like here in Australia’.
Albanese acknowledged the multitude of risks that comes with unregulated commercial AI. By seizing natural resources and intellectual property AI companies pose risks to the country’s economic, legal, social, moral and spiritual practices. Yet he offered no clear demonstration that the PM’s Office of AI will manage the complex realities of governing commercial AI interests alongside citizen rights. The actions of the government run counter to the rhetoric about protecting Australian data and sovereignty.
On April 1 the federal government signed a MOU with Anthropic. This agreement facilitates collaboration ‘on progressing the National AI Plan’s goals of capturing the opportunities of AI, spreading the benefits, and keeping Australians safe.’ Yet it is difficult to determine what exactly the opportunities are that come with commercial AI: across government websites and LinkedIn posts by AI fans, the language around benefits are woolly and abstract. Public and expert analyses in news and social media about commercial AI practices offer no evidence of opportunities.
When we consider the very tangible costs to Australian society, the ‘advantages’ of AI seem even harder to grasp. The government’s own report on AI’s impact on the labour market found that AI-enabled job displacement is slowing employment growth and will disproportionately impact female-dominated occupations. The 290 data centres that currently exist in Australia strain local energy grids and water resources.
Furthermore, the current business practice of AI companies, such as Anthropic, involves data theft. To source cultural and intellectual content – books, articles, images, songs, music, and videos – they take, without payment, enormous quantities of content from online depositories. To work around copyright restrictions, Anthropic aim to ‘destructively scan all the books in the world’ – systematically purchasing, copying, then destroying books – to train Large Language Models (LLMs). Once trained LLMs generate “new” content, such as articles, images and videos, through AI interfaces, like Claude.
In his speech, Albanese did acknowledge that Australian creators are denied revenue and acknowledgment of their work with AI companies’ theft of their labour. Holding foreign companies like Anthropic accountable for such theft through the avenue of domestic laws and industry standards would be a start. However, the practicalities of doing so were not addressed in the PM’s address.
Data sovereignty is concerned with access to personal information, consent for use, and profit; who owns, controls, and benefits from the data of an individual or community. It was recognised as a key issue by the Albanese government’s 2023 report on international digital platforms.
Data sovereignty is not a new issue. Women, Indigenous individuals and communities, queer people and people with disabilities have historically had their data used against them. Consequently, data sovereignty is now a well-established concept in Indigenous scholarship, which guards the rights of communities to govern their data.
With the evolution of digital technology, data sovereignty has become a more widespread issue. The commodification of personal digital data- such as social media posts, images and messages – that began in the 1990s is now embedded in everyday life. As scholars such as Shoshana Zuboff has shown, digital companies are built on a model of ‘surveillance capitalism’, where companies buy, sell and exploit our personal data for profit – even using data to influence and modify our behaviour. Companies that claim Australians will benefit from the introduction of AI — such as Google, Microsoft, and Meta (formerly Facebook) – are the ones who look to profit as citizens exist, professionally and personally, online. online.
Data sovereignty is thus an issue with a history that pre-dates the emergence of AI technologies. It is an issue posed particularly sharply for marginalised communities, but which has implications for every Australian. Yet the PM described sovereignty only in terms of capability, or infrastructure, and engaged with individual rights only in the narrow terms of copyright violations. Albanese could have taken the concept of sovereignty further to provide a stronger regulatory, and ethical, response to commercial AI.
The AI industry could be forced to provide individuals with the right to opt-out of using AI components at work (many of which are slowing down, rather than improving workflow); give individuals more control over who can access their digital data, which is now routinely created in education, medical visits, entertainment, and shopping; and temper the unnecessary expansion of data centres across the country. The government could also protect Australians from the copyright violations of generative AI with development of practical technologies which identify instances of data theft. These steps would be an ethical response to what is currently a profoundly unethical industry.
Furthermore, regulation is only as good as the people in the room. Albanese’s national framework of AI governance relies on the newly created Office of AI in the Department of the Prime Minister and Cabinet, which will include ministers for Industry and Innovation, Science, Technology and the Digital Economy, Climate Change and Energy, Education, Communications, Home Affairs/the Arts, Employment, the Attorney-General, and the Treasurer would be present. It is concerning that the Ministers for Women and Indigenous Affairs – two groups who have historically been disproportionately affected by violations of data sovereignty – are not represented in this Office, particularly because AI ‘mirrors the gender bias in our society’.
AI in Australia’s interests acknowledges Australians’ fears and anger concerning the intrusion of commercial AI into all areas of their lives, but so far it’s a light-on-detail framework for dealing with this critical challenge to our collective futures. More details will emerge when the National Cabinet meets. Before then, the government can work on bringing women, gender-diverse people, First Nations voices, and other experts into the PM’s Office of AI. Doing so would provide Australians with a stronger starting point by which to govern commercial AI interests.
Share