Australia proposes mandatory checks for its most advanced AI models
The government proposes independent evaluations, rapid incident reporting and a new regulator. The obligations have yet to be enacted.
On 10 November the Australian government announced plans to put mandatory safety rules into Australian AI Standards legislation. They would apply to companies seeking to offer or develop in Australia models at the cutting edge of current AI capabilities. The announcement concerns this category of model and does not automatically extend the obligations to every business that uses AI.
The plan calls for independent evaluations, security controls from development through deployment, management of serious risks and rapid notification of incidents to the government. Companies would also have to publish their safety frameworks and adapt to updated standards.
The government proposes a new regulator with the power to impose penalties, working with the Australian AI Safety Institute for technical expertise. The announcement did not publish final legislative text, specific penalties or a start date for the obligations.
English translation of the Greek edition · Read in Greek