Australia’s five new AI safety priorities put privacy law, workplace protections and consumer rights at the centre of the government’s next move on artificial intelligence. Notably, they arrive with a named minister attached to each one. The announcement, made on 20 July in a joint statement from six ministers through the Department of Industry, Science and Resources, sets out where Canberra intends to concentrate its regulatory energy under the National AI Plan released last December.
The signatories span five portfolios: Attorney-General Michelle Rowland, Industry and Innovation Minister Tim Ayres, Communications Minister Anika Wells, Employment and Workplace Relations Minister Amanda Rishworth, and assistant ministers Andrew Leigh and Andrew Charlton. That breadth is the point. Rather than drafting a standalone AI Act along European lines, the government is doubling down on its strategy of stretching existing law across privacy, consumer and workplace rules to cover algorithmic harms.
If you build, deploy or simply use AI tools in Australia, this list of AI safety priorities is the closest thing you’ll get to a regulatory roadmap for the next two years. It tells you which reviews are live, which laws are about to move, and which minister to watch for each of them.
Where the five AI safety priorities land
The first priority is a legislated digital duty of care, led by Wells, which would require technology companies to build safety into their products by design and address foreseeable risks before they reach users. It extends an obligation the government has already flagged for social media platforms into AI products. The second is privacy: Rowland will consult on a second tranche of Privacy Act reform aimed at strengthening how personal information is protected in both conventional and AI-driven services, picking up the reform program that stalled after the first tranche passed in 2024.
Workplace AI comes third. Rishworth will pursue AI safety as one of the agreed priority areas of the tripartite Artificial Intelligence Workplace and Employment Forum, where unions, employers and government are negotiating how automated systems can be introduced into workplaces without eroding conditions or oversight. Fourth, Leigh will examine options under the Australian Consumer Law to deal with two distinctly modern problems: retail surveillance pricing, where companies use personal data to charge different customers different prices, and agentic commerce, where an AI agent rather than a human does the buying. Fifth, the Attorney-General’s department will develop a framework governing how federal agencies themselves use automated decision-making, a direct legacy of Robodebt, which showed what happens when government automation runs without guardrails.
The ministers’ framing is deliberately optimistic. “Artificial intelligence is transforming the way Australians work and learn, and can deliver significant economic and social benefits when shaped and harnessed well,” the joint statement reads. “The National AI Plan makes clear that to fully realise these benefits, Australians must have confidence that AI is being developed and deployed safely.” The statement also insists the government “views safety and innovation not as competing priorities, but as complementary goals”, a line aimed squarely at the tech industry’s standing complaint that regulation strangles adoption.
The announcement leans on machinery that already exists. The AI Safety Institute, launched alongside the National AI Plan, will supply the technical evaluation work, alongside CSIRO and the Gradient Institute, and Australia has joined the International Network for Advanced AI Measurement, Evaluation and Science to share testing methods with counterpart bodies overseas. The government also points to runs already on the board, including new criminal offences for non-consensual sexually explicit deepfakes and moves against so-called nudify apps.
A list of priorities, or a strategy?
Not everyone reads the announcement as progress. Writing in The Conversation two days after the release, Henry Fraser, a research fellow in law, accountability and data science at Queensland University of Technology, argued that the messaging about “priorities” and “AI safety” is “doing a lot of heavy lifting”, noting that most of the five AI safety priorities repackage reviews and consultations that were already under way. Fraser points to the government’s “pattern of drift, fragmentation and delay” on AI policy. The privacy reforms have been in motion since 2020, and a proposed set of mandatory guardrails for high-risk AI, consulted on in 2024, has quietly disappeared from the agenda.
The resourcing question is harder to wave away. The AI Safety Institute is funded at $29.9 million over four years, according to figures cited in Fraser’s analysis, a modest sum for a body expected to evaluate frontier models. An expert advisory body established to guide the government’s AI work was wound up after 15 months, having cost just $200,000, according to the same analysis. It’s hard to escape the sense that the announcement’s six ministerial signatures were easier to assemble than a single line of new funding.
Still, attaching each priority to a named portfolio creates something the government’s AI agenda has lacked: accountability. When the second tranche of privacy reform slips, that now lands on Rowland. If the consumer law review into surveillance pricing produces nothing, Leigh owns the gap.
The next 12 months will show whether the list converts into law. The privacy consultation is the first concrete process to watch, followed by the shape of the digital duty of care legislation and whatever the workplace forum produces on consent and disclosure when employers deploy AI on staff. Businesses using AI for pricing or customer profiling should treat the consumer law review as a warning shot. Canberra has now told everyone where it intends to act on AI safety. The harder task is proving these five commitments move faster than the reviews they grew out of.
