Australia’s AI regulation trajectory: Get up to speed on the latest developments
Australia’s AI reform agenda has accelerated, and it is moving on several fronts at once. In the past week alone, the Federal Government has committed to mandatory AI Standards, accepted all 18 Rapid Review recommendations, and put hard dates on privacy and online safety reforms.
On 10 November 2026, the Federal Government announced that it will legislate mandatory AI safety and security protections in Australian AI Standards for frontier AI companies (AI Standards), to be enforced by a new AI safety regulator. The AI Standards are targeted at large international AI developers and comprise the first stage of the Federal Government's reform agenda. We anticipate the next stages of the reform agenda will impact other players in the AI lifecycle including application developers, deployers, and users of AI. The Federal Government has repeatedly foreshadowed the creation of new and uplifted laws with whole-of-economy application (e.g. through privacy reform, competition reform, digital duty of care, online safety requirements).
In short, binding AI obligations are crystallising across Australia, at the state and federal level, across various legislative instruments, and on different timelines.
In the coming weeks, significant reports and announcements will further shape expectations of good AI governance in Australia. Understanding the recently announced whole-of-economy expectations driven by the Federal Government is key, as is mapping these expectations to the broader reform agenda underway. The window to prepare and respond is short.
Mandatory AI Standards and a new AI regulator
The AI Standards, announced on 10 November 2026, apply to frontier AI companies and require them to:
- maintain security controls across all stages of development and deployment;
- identify and manage serious risks associated with their models;
- allow independent evaluation of their models;
- provide mandatory notification of incidents to the Australian Government rapidly;
- maintain and publish safety and security frameworks to ensure good governance, risk management and transparency; and
- respond to updated standards and rules as technologies change.
Compliance with the AI Standards will be enforced by a new, specialised AI regulator, with powers to impose "significant penalties" - including potential criminal liability - for breach of the AI Standards.
The Federal Government also announced that its key priorities in managing AI are to:
- provide safety for Australians and keep their data secure;
- keep national security data protected and in the national interest; and
- ensure AI companies have a social license to operate.
Recommendations from the Rapid Review into the Open AI Medicare incident
On 10 November 2026, the Federal Government's Rapid Review released 18 recommendations, which have been accepted by the Federal Government. The recommendations relate to an incident that occurred in June 2026, where without authorisation, an OpenAI agent accessed Services Australia’s Medicare Statistics Reporting Service during internal OpenAI testing.
While the full Rapid Review report has not yet been published, the recommendations, and the messages surrounding their acceptance are significant, signalling new expectations across key areas, including:
- enhanced Federal Government engagement with frontier AI developers;
- mandatory and timely AI incident reporting;
- focus on domestic and international development of AI standards;
- uplift of Government AI incident response, co-ordination and escalation procedures; and
- modernised cybersecurity practices across the Federal Government, fit for AI enabled threats.
More than half of the Rapid Review recommendations are directed at the Federal Government’s own systems and procedures. The rest look outward to AI developers and the legal framework for AI safety and security in Australia.
Recommendation 8 of the Rapid Review is particularly significant. It asks the Attorney-General’s Department to consider options to reform the criminal law to improve AI safety, in collaboration with the Australian Federal Police, Home Affairs and other agencies. The options include criminal penalties for egregious regulatory non-compliance and a standalone “failure to prevent” offence. It is foreshadowed that the offence would apply where a company responsible for developing an AI model that engages in harmful conduct fails to take reasonable steps to prevent that harm. If enacted, it would in effect require AI developers to take reasonable steps to prevent harm caused by their models. This, in turn may have implications across the supply chain.
Frontier AI developers would effectively have a positive duty to mitigate harm while developing AI technology in Australia, failing which, they could face criminal sanctions. What constitutes reasonable steps in the context of development of novel AI technologies is not yet clear. Nor is it clear how this new proactive harm mitigation requirement will interact with other announced laws, such as the digital duty of care.
JSCAI open letter to Prime Minister - 12 October 2026
These announcements come against the backdrop of the JSCAI's inquiry. Established by both Houses of Parliament on 20 August 2026, the JSCAI is examining the economic, legal, security and social dimensions of AI and its impact on Australia.
On 12 October 2026, through a letter signed by its Chair and Deputy Chair, the JSCAI took the unusual step of publishing a letter to the Prime Minister, outlining early findings and recommendations from its inquiry. The letter:
- called for safeguards for frontier AI developers deploying models and agents in Australia. These would cover oversight and accountability, safety and risk management, incident monitoring and reporting, independent evaluation, protection of Government data, and international interoperability covered by standards.
- highlighted Australia’s copyright system was not facilitating licensing deals for Australian culture and content;
- recommended that nationally consistent data centre standards be developed; and
- recommended principles for the development and deployment of AI to build sovereign capability. These would cover data sovereignty, critical AI infrastructure and energy, AI literacy, skills and procurement.
The JSCAI’s final report is due by 30 November 2026 and will likely inform how the AI Standards are drafted, and which areas are prioritised. The report may also shape obligations beyond those for frontier AI developers, including, for example, for deployers in high-risk settings.
Other developments
- Automated decision-making requirements: New transparency obligations come into force 10 December 2026, requiring the disclosure of information in privacy policies regarding automated decision-making which involves personal information.
- Digital duty of care reforms: New laws are progressing through Parliament requiring online services, including games, social media platforms and generative AI services, to take reasonable steps to ensure a safe online environment for Australians, with specific protections for children. See our article: Australia's new digital duty of care: What it means.
- Australian Standards for AI and data centres: The Federal Government plans to introduce mandatory standards for data centres and frontier AI training, with tiered obligations for data centres based on power capacity covering areas such as new generation underwriting, connection costs, water use, and facility location.
- Safety in the workplace and new consumer law protections: As referenced in the Assistant Minister for Science, Technology and the Digital Economy’s 8 October 2026 Speech, the Minister for workplace and employment is looking at safety in the workplace and the Assistant Minister for Competition and Consumers is exploring issues to do with agentic commerce and retail surveillance pricing.
We set out more detail on automated decision-making requirements, the digital duty of care, and AI and data centre standards in the table below.
Implications
While several key issues remain unsolved, expectations for AI governance are rising. Taken together and considering the broader landscape of AI reform below, we recommend closely monitoring the key milestones below, and taking the following precautionary actions now:
- Develop a current inventory of use cases and data inputs to understand your AI footprint. This is the foundation for automated-decision transparency obligations, duty-of-care risk assessments and future standards compliance. The inventory should consider AI use across the supply chain and how standards for frontier AI companies may impact your organisation.
- Review supplier arrangements and confirm where AI providers host and process data, what requirements exist in regard to incident notification, model changes, and regulator engagement.
- Meet the 10 December 2026 deadline for the automated decision-making transparency obligation, confirming which automated decisions are in scope and updating privacy policies accordingly.
- Assess online safety exposure and determine whether any consumer-facing AI service triggers the incoming digital duty of care.
- Engage now, including considering any submissions to upcoming consultation on Australia's AI Standards, as well as South Australia's Royal Commission into AI by 30 November 2026.
Key dates
| Date |
Milestone |
| 20 November 2026 |
Senate committee to report on the Digital Duty of Care Bill |
| 30 November 2026 |
Joint Select Committee final report; SA Royal Commission submissions close |
| 10 December 2026 |
Automated decision-making obligations commence; Children’s Online Privacy Code registered |
| End of 2026 |
Australian Standards for AI expected |
| 27 March 2027 |
Environment and Communications References Committee to issue final report on AI and data centres |
| Early 2027 |
AI Standards legislation to be introduced |
| 1 July 2027 |
South Australian Royal Commission to release final report |
Recent developments
The Joint Select Committee on AI (JCSAI)
Description: The JCSAI was established on 20 August 2026 by resolution of the House of Representatives and Senate. Its terms of reference cover productivity, data sovereignty and domain-specific models trained on Australian data, workforce and skills, the adequacy of existing laws, and copyright. They also cover scams, deepfakes, risks to children and vulnerable groups, and national security.
On 12 October 2026, the JCSAI took the unusual step of writing to the Prime Minister with early advice, which it described as an indicator of the direction of its work. The letter’s centre of gravity is frontier AI. It recommends:
- a risk-based framework covering model evaluation, safety testing, transparency and human oversight;
- a duty on frontier labs to monitor and report incidents within clear deadlines;
- independent evaluators embedded within frontier labs; and
- early access to new models for security agencies, as a condition of training in Australia.
The JCSAI also recommends a standing parliamentary committee to oversee AI regulation and publish details of incidents involving unauthorised actions by AI agents. Beyond safety, it calls for requirements governing where data in AI systems is stored and processed and who can access it. It also calls for procurement rules favouring Australian providers, a register of sovereign AI companies and a review of copyright settings.
Implications for developers and deployers: For frontier developers, the direction is clear. Time-bound incident reporting and independent evaluation are likely to become conditions of operating in Australia. Detection and escalation processes should be tested against a reporting deadline now.
For most other organisations, what is left out of the early advice matters more. Key terms of reference remain open, and the final report may contain recommendations aimed squarely at them. These include consumer protection, scams, deepfakes and gaps in existing law. In the meantime, focus on sovereignty requirements has increased scrutiny of supply chains: where AI services are hosted, who can access the data and what vendor contracts permit. The JCSAI also prefers regulations over primary legislation, so obligations may change faster than businesses are used to.
South Australia’s Royal Commission into AI
Description: The Royal Commission commenced on 1 October 2026 and must report by 1 July 2027. Its Scoping Paper, released on 29 October 2026, proposes six themes:
- AI in society;
- better services;
- growing innovation;
- managing the economic transition;
- education and skills; and
- safe and trusted AI.
Data centres are excluded, as are matters primarily for the Federal Government, including intellectual property and copyright. Submissions close on 30 November 2026 and will be published by default.
Implications for developers and deployers: The Commission is policy-focused, but its recommendations are likely to influence thinking well beyond South Australia. They will centre on the levers the state controls, and procurement is likely to feature prominently. Suppliers to the South Australian Government should expect future contracts to place greater weight on transparency, audit rights, data location and incident notification. Other likely areas of focus include:
- AI and automated decision-making in state and local Government;
- data governance (South Australia has no public sector privacy statute);
- workforce protections; and
- AI’s effect on children.
Organisations considering a submission should remember it will be published and must be consistent with positions taken before other regulators. Suppliers should also note that a Royal Commission has compulsory powers to obtain documents, including from the agencies it examines.
Australian Standards for AI and data centres
Description: National Cabinet affirmed the Federal Government’s plan for mandatory AI and data centre standards on 26 August 2026. The standards are expected by the end of the year, with legislation in early 2027. The September consultation paper indicated the standards will cover minimum requirements for data centres and conditions for frontier AI training in Australia.
For data centres, the paper proposed three tiers. Facilities under 30MW would sit outside the standards. Facilities between 30MW and 100MW would face baseline requirements, with more substantial obligations above that. Operators are expected to:
- underwrite new generation;
- meet their full share of connection costs;
- minimise water use; and
- locate facilities appropriately.
These obligations will sit alongside existing state planning approvals rather than replacing them.
On frontier training, the Federal Assistant Minister for Science, Technology and the Digital Economy has described a “systems-based” approach drawing on banking supervision, aviation and workplace safety. Developers of the highest-risk models would need to show that their processes for managing risk actually work.
Implications for developers and deployers: The data centre standards will apply directly only to operators of large facilities. For deployers, the effect will be indirect but real. The standards will shape the cost and availability of onshore compute, and so how practical it is to meet data residency expectations. For model developers, the training conditions are the space to watch. Being able to evidence a working safety management system is likely to become a regulatory expectation, and it will become a standard question deployers ask of their providers.
Digital duty of care
Description: The Online Safety Amendment (Digital Duty of Care) Bill 2026 was introduced on 15 October 2026 and passed the House of Representatives on 28 October 2026. A Senate committee is due to report on 20 November 2026.
As we outlined in Australia’s new digital duty of care: What it means, that the Bill replaces the Basic Online Safety Expectations and industry codes with a single duty. That duty requires responsible persons to ensure, so far as is reasonably practicable, a safe online environment. It expressly captures generative AI services. It can reach not only named providers but also those exercising day-to-day control over a service, including underlying technology providers.
Responsible persons must prepare written risk assessments at least annually. The Bill would also bar AI chatbots from building emotional intimacy with children. Penalties reach approximately $109 million, and the core duty commences 12 months after Royal Assent.
Implications for developers and deployers: The duty marks a significant shift from reactive, incident specific and narrow compliance obligations to proactive compliance obligations requiring persistent identification and mitigation of risk.
- Developers of consumer-facing generative AI are squarely in scope. They should use the transition period to prepare their first risk assessment, with particular attention to age assurance and children.
- Organisations embedding chatbots or generative features in public-facing services should assess whether they could be a “person responsible”. Where they rely on a third-party model, they should establish what safety commitments the provider will make in the contract.
eSafety’s guidelines will not provide a safe harbour. A reasoned record of what “reasonably practicable” means for a given service will matter more than generic compliance.
Privacy and automated decision-making
Description: From 10 December 2026, transparency obligations come into effect which require certain information to be included in privacy policies if an entity uses or arranges for a computer program to use personal information to make decisions that could significantly affect individuals. The obligation applies to decisions made from that date, even if the system was established earlier. Using a vendor system does not shift due diligence and disclosure obligations to the vendor.
The Children’s Online Privacy Code must be registered by 10 December 2026, although its commencement date has not been confirmed. Further ahead, a key feature of the Tranche 2 exposure draft released on 31 August 2026 centres on a new “fair and reasonable” test for handling personal information. For detailed analysis of these reforms, see our articles here: OAIC's Children's Online Privacy Code: what to expect - Technical update; Draft Tranche 2 Privacy Reforms: What they mean for businesses - The highly anticipated draft Tranche 2 privacy reforms have arrived.
Implications for developers and deployers: The automated decision-making transparency obligation is the most immediate deadline in this update. It applies to any organisation using computer programs (including AI) to make, or substantially assist with, significant decisions about individuals, such as in recruitment, lending, insurance or eligibility for services. Deployers should identify those decisions and update their privacy policies before 10 December 2026.
The scrutiny these systems attract is real. In October 2026, the Commonwealth Ombudsman found that an automated aged care assessment tool lacked meaningful transparency and risk assessment.
Developers should expect customers to ask for information required to comply with privacy obligations. They should also consider how a “fair and reasonable” test would apply to training AI on personal information.
Final thoughts
For boards and executives, the challenge is understanding how various obligations, reforms, and policy positions intersect across various timelines, and preparing AI governance and strategy in response.
The pace of regulatory change has accelerated as Australia prioritises and fast tracks AI reform. Engagement with the Federal Government and regulators should be a priority for impacted businesses that wish to ensure their voices are heard in this period of rapid change. We anticipate that the next few years will follow a similarly fast paced cadence as AI technology continues to mature into a wider array of applications and industries.
Noting the key dates mentioned above, the coming weeks are particularly significant.
If you would like to discuss how these developments affect your organisation, including preparing for upcoming automated decision-making transparency requirements or a Royal Commission submission, we can assist.