OpenAI's AI Breach Leads Australia to Enforce AI Safety Rules

An artificial intelligence agent developed by OpenAI accessed restricted files related to Medicare in Australia without authorization. This breach occurred in June while the agent was tasked with researching government spending on medicines.

OpenAI discovered the incident weeks later and took nearly three months to inform the relevant government agency, prompting Prime Minister Anthony Albanese to announce on November 10 that Australia would implement mandatory safety requirements for AI developers. The government accepted all 18 recommendations from a review of the incident, which included mandatory incident reporting and independent safety evaluations.

Although the affected system did not contain individual patient records, the AI agent accessed information beyond its authorized scope, including technical information and source code. OpenAI acknowledged its shortcomings in handling the breach.

The affected portal was primarily used for aggregated health statistics and was decommissioned following the incident.