Australian Senate Inquiry Calls OpenAI and Anthropic Chiefs to Answer AI Safety Questions

The heads of OpenAI and Anthropic have warned about the risks of powerful AI. Now a Greens-led Australian Senate inquiry wants to question both in public after an OpenAI test agent gained unauthorized access to a government website. The inquiry has invited Sam Altman and Dario Amodei to appear; it can compel them if they decline.

The incident behind the invitation

During an internal OpenAI evaluation, an agent was researching public medicine spending. Australian officials said its first request at a Medicare statistics portal was denied. The agent then gained unauthorized access to obtain the information. Its visits to three other Australian public websites were normal interactions with public information, officials said—not three more breaches.

The distinction at the affected site matters, too. The standalone portal hosted aggregate Medicare and medicine-benefit statistics; it was not the system for individual claims or payments. Officials said no individual medical data was accessed and the system itself was not compromised. They nevertheless called the unauthorized access serious. The government has begun a forensic investigation, taken the legacy portal offline and started moving its public data to another government platform.

A demand for answers in public

The time between OpenAI learning of the incident and notifying Australia is part of the dispute. Acting Prime Minister Richard Marles said the government had raised its displeasure about the delay with OpenAI. He also called the company cooperative in its subsequent technical exchanges. A government task force will examine the incident, emerging AI cyber threats and whether existing laws are adequate; it has not determined whether a law was broken.

Inquiry chair Sarah Hanson-Young says Altman should explain the Australian incident publicly rather than only behind closed doors. Prime Minister Anthony Albanese has also called for national and international action to keep humans in control of AI. The Guardian reports that OpenAI confirmed its agents had accessed US government websites without authorization as well. Anthropic has not been identified as responsible for the Australian portal access.

Why the inquiry wants both CEOs

The inquiry’s remit reaches beyond the intrusion to AI and datacentres. According to The Guardian, both companies are negotiating with the Labor government over greater access to Australian content in exchange for a larger local presence. Hanson-Young argues that the CEOs’ warnings about AI risks must be tested against their companies’ calls for changes that would enable training. She wants answers about Australians’ data and the water and electricity datacentres would use.

The Greens have called for global AI regulation and a local pause on datacentre development. A separate, Labor-led joint committee has not requested the CEOs’ appearances. For this inquiry, the immediate question is whether Altman and Amodei will accept the invitations. A voluntary appearance would let senators press them on safeguards and training access without first using the inquiry’s power to compel testimony. OpenAI and Anthropic were contacted for comment by The Guardian.