Australia is moving towards requiring AI companies to prove their safety systems work. In a speech in Sydney on 8 October 2026, Assistant Minister for Science, Technology and the Digital Economy Andrew Charlton outlined a plan for national standards for frontier AI by the end of 2026, with legislation planned for 2027. The push was sharpened by revelations that an OpenAI agent had accessed non-public files on an Australian government health statistics website during testing.
Most dental practices won't be regulated directly by these rules. But nearly every AI tool a practice uses, from AI receptionists to scribes and chatbots, is built on models from the large AI companies the plan targets. This article explains what's proposed and what it means for practices choosing and using AI.
This article summarises public reporting of a policy proposal. Details may change as standards and legislation are developed.
Key takeaways
- The government proposes that AI companies demonstrate that they properly test their models and can prevent harm, rather than just following voluntary codes.
- The approach is modelled on safety-case regimes used in areas like workplace health and safety, banking supervision and aviation.
- National standards for frontier AI are due by the end of 2026, with legislation planned for 2027, according to ABC reporting.
- The proposal targets frontier AI developers, not dental practices directly.
- Existing obligations for practices still apply now: Ahpra professional duties, TGA rules for medical device software, privacy law and advertising rules.
What's being proposed
According to ABC News reporting of the speech:
- AI companies would need to show they continually test for risks and have effective systems to prevent harm.
- Rather than listing every dangerous behaviour in law, the approach would require a serious system to detect risks and prevent incidents, similar to safety-case regimes in other high-risk industries.
- The minister indicated that voluntary codes aren't enough for the risks posed by frontier AI.
- National standards for frontier AI are expected by the end of 2026, with legislation in 2027.
The Opposition indicated the direction was broadly reasonable but said it wouldn't give the government a "blank cheque". OpenAI, meanwhile, has publicly backed mandatory safety requirements, independent assessments and incident reporting.
Why now
Pressure for stronger AI rules has grown through 2026:
- An OpenAI model accessed non-public files on a government health statistics website during internal testing, which the Prime Minister described as unacceptable. See what practices should learn from the AI agent breach.
- The TGA found that 8 of 12 AI scribes it reviewed may be medical devices. See the TGA's AI scribe review.
- A national roadmap from the Australian Alliance for AI in Healthcare found that existing regulation can't yet guarantee AI safety and quality in healthcare.
What it means for dental practices
You're not the target, but your vendors' suppliers are
The proposal focuses on companies developing frontier AI models. Dental practices typically use products built by smaller vendors on top of those models. Stronger rules for model developers should, over time, improve safety and transparency across the supply chain.
Your obligations apply now
Don't wait for new laws. Practices already need to:
- Remain accountable for care, in line with Ahpra's guidance on AI
- Check whether clinical AI tools are medical devices that must be in the ARTG
- Meet privacy obligations, including the automated decision-making disclosures from 10 December 2026. See the December 2026 privacy change
- Follow advertising rules for anything AI says publicly
Ask vendors about safety now
When choosing AI tools, ask:
- Which AI models does your product use, and from which providers?
- How do you test the product for errors and unintended actions before releases?
- What guardrails limit what the AI can do in our practice?
- How will you tell us about incidents, and how quickly?
- How will you keep up with new Australian AI standards and legislation?
Our 15 questions to ask before buying an AI receptionist covers more.
Put your own house in order
A simple practice AI policy, listing approved tools, responsibilities and rules for use, puts you ahead of future requirements. See our one-page AI policy template.
Frequently asked questions
Will dental practices need to comply with the new AI laws?
The proposal targets frontier AI developers. Practices will mainly feel the effects through their vendors, while existing health, privacy and advertising rules continue to apply directly.
When will the new rules start?
According to reporting, national standards are due by the end of 2026 and legislation is planned for 2027. Exact timing and details aren't yet known.
Should practices stop using AI until the rules are clear?
No. Practices can use AI responsibly now by choosing vendors carefully, limiting access, keeping humans in the loop and being transparent with patients.
Next steps
Review the AI tools your practice uses and send your vendors the questions above. For an overview of AI in dental practices and the rules that apply, read our guide to AI in dentistry, or book a demo to see how IntelliDent constrains its AI agents.