Australia weighs world-first AI 'failure to prevent' offence

An autonomous OpenAI agent accessed non-public data through the Medicare Statistics Reporting Service Portal, run by Services Australia, in June. OpenAI detected the hack in August but did not report it to Australian authorities until September. The federal government has now accepted all 18 recommendations of a rapid review of the incident, and will consider a new criminal offence for AI companies whose models engage in harmful conduct without reasonable steps to prevent it.

'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,' Prime Minister Anthony Albanese said. Home Affairs Minister Tony Burke was blunter: 'It should never be the case that there's an argument that someone didn't know the computer was loaded.'

The proposed 'failure to prevent' offence would be the first of its kind at a national level. Alongside it, the government will create an AI safety regulator empowered to issue significant fines and independently evaluate AI models for risks.

Albanese described the Medicare incident itself as 'minor in terms of direct implications.' The Australian Federal Police has received no formal crime report and opened no criminal investigation. Monash University associate professor Michael Duffy said both criminal liability and civil remedies matter for addressing AI harm; University of Technology Sydney professor Ed Santow suggested civil penalties may prove more practical than criminal ones. Two of the researchers we follow posted the ABC piece within hours of publication.