$ 44.85 € 50.23 zł 11.47
+12° Kyiv +15° Warsaw +22° Washington

OpenAI admits mistakes after breaches of Australian government resources — BBC News

Fedir Kryshtovskyi 06 October 2026 09:11
OpenAI admits mistakes after breaches of Australian government resources — BBC News

OpenAI has acknowledged that its response to an agent operated by the company gaining access to Australian government resources was insufficient. OpenAI Chief Strategy Officer Jason Kwon said this during parliamentary hearings on artificial intelligence in Sydney, BBC News reports.

According to Kwon, the incident should not have happened, and the company should have responded to it better. The Australian side was informed only several weeks later, through an email sent to a general mailbox. Kwon acknowledged as a mistake that OpenAI representatives did not contact ministers directly immediately after learning about the breach.

Access to the Medicare portal

In June, an OpenAI agent, according to the BBC, gained access to a closed statistical portal containing non-sensitive data from Medicare, Australia’s universal health insurance system. Cybersecurity experts called it the first case of its kind.

Kwon said OpenAI has changed its procedures for similar incidents. The company plans to notify the affected party and begin working jointly on the situation even if it does not yet have a full understanding of the circumstances.

More current news is available on the UA.News Telegram channel Telegram.

New safeguards during testing

OpenAI has also added additional security measures to environments where it tests its models. According to Kwon, models are now monitored in real time during testing, and the system raises an alarm if they interact with the internet in an unintended way.

According to him, this enabled OpenAI to notify the authorities of New South Wales about another unauthorized access incident last week within 48 hours. The company is also creating a local working group in Australia to find ways to better manage the risks associated with increasingly powerful artificial intelligence systems.

Anthropic review

Anthropic’s head of security, Dave Orr, said at the same hearing that the company had analyzed hundreds of millions of interaction records in search of possible similar breaches. According to him, no cases of access to Australian resources were found.

Read us on Telegram and Sends

Download our app