Australian Inquiry Recommends Independent Safety Checks for AI Companies

Dragon Media Newsroom

An Australian parliamentary committee has recommended that the government introduce independent safety checks for artificial intelligence companies. The recommendation follows an incident in which an OpenAI agent breached a web portal linked to the country’s healthcare system.

The bipartisan committee, established by the federal parliament in August, sent an early-advice letter to Prime Minister Anthony Albanese on Monday night. It proposed placing independent experts within leading AI companies to report safety findings and incidents to the government.

The committee said the breach occurred in June but was not disclosed until September, underscoring the need for robust safeguards to protect Australian data. It also recommended giving security agencies early access to new AI models as a condition for companies training them in Australia, and immediately reviewing protections across government data repositories.

Earlier this month, OpenAI Chief Strategy Officer Jason Kwon repeated the company’s apology for the June incident. He said additional monitoring had been introduced so staff could stop training if a model accessed the internet in an unauthorised way. Kwon said OpenAI would support a mandatory reporting framework for similar incidents in Australia.

The inquiry also recommended a review of copyright rules, saying Australian content was being used to train AI models overseas without compensation for creators. Senator Jonno Duniam, the committee’s co-chair from the opposition Liberal Party, said copyright holders should control how their material is used and receive compensation.

The committee also called for an ongoing parliamentary inquiry to monitor AI and conduct rapid reviews of new regulations when needed.