Today, the Australian government announced its intention to include mandatory safety and security requirements in artificial intelligence standards legislation, targeting companies developing advanced models that wish to operate within the country, with a proposal to establish a specialized regulatory body to monitor compliance with these requirements and impose significant financial penalties on violators.
Proposed commitments include maintaining security controls during the development and deployment phases of models, identifying and managing critical risks, enabling independent evaluation of models, prompt reporting of incidents to government, publishing safety and security frameworks, and complying with updated standards as technology evolves.
The announcement came following a government review of an unauthorized access incident carried out by an artificial intelligence agent affiliated with the company “OpenAI” to the healthcare statistics portal “Medicare” last June during internal tests, where it was able to access files and information within the system.
The company confirmed that individual patient and customer records were not compromised, and apologized for the incident and for the delay in notifying the authorities.
At the same time, the government announced its acceptance of the 18 recommendations concluded by the review, including imposing mandatory reporting of incidents related to the behavior of artificial intelligence systems, enhancing information exchange and responding to cyber incidents, and studying amendments to criminal laws in cases of serious violations.
The government confirmed that the new requirements will be included in the artificial intelligence standards legislation, the details of which are still under preparation and consultation, and are scheduled to be presented to Parliament during the next year.