Four Priorities for the New Office of AI
Australia is seeking to navigate the tricky path of taking advantage of artificial intelligence without succumbing to its dangers.
The Prime Minister’s recent address at the University of Sydney, ‘AI in Australia’s Interests’, highlighted the establishment of the Office of AI within the Department of the Prime Minister and Cabinet.
At Synlaw, we’ve always maintained that AI shouldn't just make us faster; it should make us better. But being ‘better’ requires guardrails to ensure Australia’s interests, including both productivity and security, are upheld. The Office of AI must be proactive and ensure Australia’s uptake of AI maximises this national interest.
By placing the Office of AI directly within the Department of the Prime Minister and Cabinet, the Government has signalled that AI is not isolated to just a digital issue - it is a matter of national sovereignty and economic architecture. Whilst we await more detail as to what the new office will actually do, here are four areas that Synlaw believes would be critical to prioritise.
Adopting a Risk-Based Regulatory Framework
In his speech, the Prime Minister noted that Australia aims to be the first country to bring AI issues into a single national framework. To do this effectively, the Office should assess the global landscape and learn from the EU AI Act, which follows a risk-based approach. The Australian standards should be developed to distinguish between ‘unacceptable’ practices that are prohibited and ‘high-risk’ systems that require strict mandatory requirements regarding data quality, technical documentation, and human oversight.
For ‘high-risk' systems, compliance cannot be a voluntary ‘tick-box’ exercise. This is to ensures that innovation in low-risk settings can continue while protecting rights in sensitive areas like automated decision-making.
High-level principles will not be enough for businesses trying to remain compliant. The Office of AI should translate overarching national standards into actionable risk tiers. The Office should enforce these tiers initially through measures such as government procurement, mandating where appropriate such that no Commonwealth or State department can procure AI systems that lack verified risk-tiering and audit logs.
Owning Our Own Systems and Accountability
A common danger when adopting AI is pretending that responsibility can be handed off to third parties. While foreign firms can supply the technology, the government cannot delegate its duties to the private sector. Australia must be careful not to subcontract its future.
The Office of AI should aim to create a conducive environment for domestic development and encourage reducing dependence on overseas technology, striking a balance between safeguarding Australia’s interests and encouraging investment.
Setting mandatory Australian Standards that provide appropriate guardrails for the investors and companies within Australia requires consideration of more than just capital investment. Sovereign data and cybersecurity standards should be high on the list as well, with measures such as mandating security audits for high-impact AI infrastructure (including domestic data centres) to ensure sensitive citizen and government data remains onshore and secure against foreign interference.
The Prime Minister’s requirement for data centres to be net-renewable energy generators creates a leverage point. Approvals for critical infrastructure such as land, power, and water access, could be tied to sovereign data compliance. If a foreign firm wants access to Australia's renewable grid to run AI workloads, mandatory data residency, local security auditing, and transparent governance must be the price of entry.
Protecting the ‘Judgment Layer’
AI is a powerful accelerator, but it is not self-aware and lacks the business and regulatory context that only people provide. Recent guidance for UK lawyers suggests that failing to make use of AI could eventually be seen as a breach of professional duty where it could improve work. However, misusing it (such as submitting AI-hallucinated legal filings), is equally dangerous.
The Office of AI should lead the way in development of standards for ‘human-in-the-loop' review, ensuring that AI never assumes responsibility for decisions and that operators retain the agency to challenge AI outputs.
Crucially, technology cannot become a liability shield for decision-makers. The Office of AI must work with regulatory bodies to establish clear legal standards of care for AI integration. If AI outputs are relied upon without documented human verification, liability must remain with the human operator.
Safeguarding Intellectual Property
Guarding national interests also means protecting local creators, publishers, and businesses whose IP powers generative models. The Office of AI should ensure that a practical, standardised licensing mechanism is developed that protects Australian artists and media organisations while providing tech developers with a legal pathway to train models on domestic content. Resolving this copyright friction ensures that Australia fosters innovation while ensuring local creators are fairly compensated for their IP.
These protections must be fair and reasonable. Asking tech developers to pay excessively for content will only reduce technology development and investment in Australia, leading to lower sovereign protections and possibly slower usage of productivity enhancing technologies. Getting the balance right will be a key challenge for government over the coming 12 months.
Our View
Traditional legal and governance services were not built for this pace of change. The Office of AI represents a new future for Australian governance, one that recognises that while the cockpit is more advanced, the role of the pilot is more essential than ever. By focusing on these four priorities, the Office can ensure that AI doesn't just make us faster, it makes us better, and safer.