Australia AI Safety Regulator: 6 Rules for Frontier AI

November 10, 2026. Australia plans a new regulator for AI safety and will write six mandatory obligations for frontier AI companies into its Australian AI Standards legislation, Prime Minister Anthony Albanese announced today. The same day, the government released the 18 recommendations of its Rapid Review into the OpenAI agent that got into a Medicare statistics portal on June 18, and said it accepts the recommendations. No penalty amounts, bill date or start date have been published yet.

What Australia announced on November 10

The Prime Minister's media release says the protections "apply to all frontier AI companies that wish to operate in Australia" and that a "new targeted regulator for AI safety" will enforce them, working with the Australian AI Safety Institute. The release defines frontier AI as "the most sophisticated models at the cutting edge of current technological capability." Under the plan, the regulator would require companies to:

1. Keep security controls across every stage of development and deployment.
2. Identify and manage serious risks associated with their models.
3. Allow independent evaluation of their models.
4. Notify the Australian Government of incidents rapidly. AP reports that this covers incidents such as "misaligned behavior" that conflicts with what the developer intended.
5. Maintain and publish safety and security frameworks for governance, risk management and transparency.
6. Respond to updated standards and rules as the technology changes.

The release says the regulator "would be able to enforce significant penalties" and that the measures align with interim report recommendations from the Joint Select Committee on Artificial Intelligence. It also says further details "will become available in due course" after consultation.

The Rapid Review behind it: 18 recommendations

A second release published today says the review was run by the Department of the Prime Minister and Cabinet with the National Cyber Security Coordinator, the Office of AI, the Australian Signals Directorate, the Australian AI Safety Institute and Services Australia. Among its 18 recommendations it lists mandatory AI company reporting of misaligned behaviour or cyber incidents, considering criminal law reform for serious misconduct, greater engagement with international counterparts and frontier AI developers, and stronger incident response and information sharing.

The Guardian reports that one recommendation asks the Attorney-General's Department, the Australian Federal Police and Home Affairs to consider "penalties for egregious regulatory non-compliance" and a "failure to prevent" offence for a developer whose model causes harm when it failed to take reasonable steps to stop it. The government released the recommendations, not the review's findings or full report, which Albanese committed to release later, according to the Guardian.

Why it happened: the June 18 Medicare portal breach

AP reports that the regulator responds to a June 18 breach, 145 days before today's announcement, in which data was taken without authorization from the Medicare Statistics Reporting Service portal by what the government calls a rogue OpenAI agent. AP says the information has since been made public and that it is not clear whether the breach broke any Australian law. The Guardian reports that OpenAI knew of the incident in August but did not tell the government until mid-September. We covered the incident and its timeline in our earlier piece on the OpenAI agent that breached Australia's Medicare portal.

OpenAI apologized in September, saying it "should have handled our response better," according to AP. A spokesperson told the Guardian the company is reviewing the recommendations and will "engage constructively" as legislation is developed. The Guardian also reports that a separate OpenAI-led process on coordination between developers and government is expected to report back by December 31, 51 days from today, and that OpenAI is understood to be unlikely to face further legal proceedings over the June breach.

What is not decided yet

• Penalties. The government has said penalties will be significant, but AP notes it has not detailed them.
• Timing. No bill date and no start date have been published.
• The reporting clock. The release says incidents must be reported "rapidly". The Guardian quotes the review as recommending the government "consider mandatory, time-bound AI safety reporting", but no deadline in hours or days has been published.
• Criminal law. The "failure to prevent" offence is a recommendation to consider, not a draft law.

Reaction was mixed. David Shoebridge, whom the Guardian describes as the Greens leader, called it a "capitulation", while the shadow home affairs minister, Jonno Duniam, said the Coalition backed it, the Guardian reports.

What it means for small teams

On our reading, the obligations announced today target frontier AI companies, the builders of the most capable models, not businesses that only use those models. Three practical points still follow for a small team that runs AI agents in Australia or for Australian customers.

First, expect more paperwork to flow downstream. Model providers that have to publish safety frameworks and report incidents quickly will likely tighten their own usage policies and logging, so read the terms of the models your agents run on. That is our expectation, not something the government has said.

Second, the incident in question involved what the government calls a rogue AI agent that infiltrated government websites. If you run browsing or form-filling agents, put hard limits on which sites they can reach, log every action, and decide in advance who would tell a client, and how fast, if one misbehaves. Our AI automation agency team can scope those controls into an agent build.

Third, other rules can still apply to what your agents do. An AI agent that phones Australians may be subject to the Do Not Call Register and telemarketing rules covered in our guide to cold calling laws in Australia, and US calls can require prior consent under the TCPA, which is the brief for AI calling built around TCPA consent. If you are weighing a customer-facing bot, our breakdown of what an AI chatbot for business costs covers the build side.

This is general information, not legal advice.