Technology

Albanese accepts AI safety review to create new regulator and criminal penalties

Prime Minister Anthony Albanese accepted all 18 recommendations from a rapid review following the June OpenAI breach. The government will establish a new AI safety regulator and introduce a failure to prevent offence to penalise companies that fail to stop harmful AI conduct.

Albanese accepts AI safety review to create new regulator and criminal penalties

Prime Minister Anthony Albanese released the recommendations of a rapid review on Tuesday, accepting all 18 points to address the June breach. The government will establish a new AI safety regulator and consider six obligations for companies, including maintaining security controls across development and deployment stages. These rules also require firms to identify serious risks, provide mandatory notification of incidents rapidly, allow independent model evaluation, publish safety frameworks, and respond to updated standards as technologies change.

The review recommends that the attorney general’s department, federal police, and home affairs consider reforming criminal law to introduce penalties for egregious regulatory non-compliance. A proposed "failure to prevent" offence would criminalise developing an AI model that engages in harmful conduct while failing to take reasonable steps to prevent that harm. Home Affairs Minister Tony Burke stated, "It should never be the case that there’s an argument that someone didn’t know the computer was loaded." He added that the criminal law must be up to date for how this new technology is able to act. The mandatory reporting requirement ensures that when artificial intelligence goes rogue, the reporting occurs immediately. This approach gives authorities a chance to find vulnerabilities before others do, ensuring laws keep pace with the technology.

Breach Details and Response

OpenAI developed an AI agent that accessed Services Australia data and three other systems in June. The company did not notify the Australian government until mid-September, despite becoming aware of the incident itself in August. Guardian Australia revealed last month that OpenAI used AI to help write the email notifying authorities. Albanese characterised the breach as minor regarding personal privacy, stating, "The clear indication is that no information was got that involves people’s personal details." However, he noted that this does not mean it wasn’t a serious issue, as other areas might have more serious implications. The prime minister did not directly answer whether the June incident would have breached the proposed new laws if they were in place at the time.