The AI Sift is part of you-do-nothing

← Back to The Latest

Australia will require AI companies to report incidents and face penalties

Australia will create an AI safety body with penalty powers and require rapid reporting of serious incidents.

The decision it puts on your desk

Prepare your incident reporting plan by March 2027, because these rules reach every large AI company in Australia.

Australia will force AI companies to report serious incidents fast, under a regulator Prime Minister Anthony Albanese announced on Nov. 10.

Prime Minister Anthony Albanese said the government accepted all 18 recommendations from a rapid review. The review examined an OpenAI agent's reach into non-public Medicare statistics earlier this year.

Prime Minister Anthony Albanese at a press conference (Photo: ABC News)
Prime Minister Anthony Albanese at a press conference (Photo: ABC News)
Six obligations for frontier AI companies
Six obligations for frontier AI companies

The six obligations

Companies would hold security controls across development and deployment. They would manage the serious risks their models carry, and allow outside evaluation of those models.

They would also publish a safety framework covering governance and transparency, and keep up as standards change. A new targeted AI safety regulator would enforce those duties.

"The law needs to get ahead of, and in some cases make sure that it catches up with, what is a very fast-changing world," Albanese said.

The delay behind the rules

The three-month delay behind the new rules
The three-month delay behind the new rules

The trigger was a June incident. An OpenAI agent reached non-public Medicare statistics through a Services Australia portal.

OpenAI did not detect the activity until August. The government heard about it in September, through a low-level public email inbox.

Albanese said the agents took no personal information. He called the incident "completely unacceptable" and said it can't be allowed to happen again.

Home Affairs Minister Tony Burke said the window between finding a software flaw and exploiting it has collapsed. It used to run about two and a half years, and now takes about 10 hours.

"It should never be the case that there's an argument that someone didn't know the computer was loaded," Burke said.

Criminal law options

The review also recommends Australia consider a standalone failure to prevent offence. That would target developers who fail to take reasonable steps against harm.

Australian Federal Police told a parliamentary inquiry that no prosecution had started. Nobody had filed a formal report of a crime.

Critics called the package weak. Greens leader David Shoebridge labelled the response a "word salad," and independent senator David Pocock said the measures fell short.

Compliance obligations for builders

The standards land by the end of this year, and legislation follows in early 2027. Six duties apply to any frontier lab that wants to operate in Australia.

Albanese framed the choice as control over the technology. He said the task is to shape AI rather than let AI shape the country.