Australia’s Senate inquiry into artificial intelligence has asked the chief executives of OpenAI and Anthropic to appear at a public hearing, extending parliamentary scrutiny of an AI agent that accessed federal government systems, including Medicare.
Written requests were sent to Sam Altman and Dario Amodei ahead of the inquiry’s next Canberra hearing on October 1. They are invitations to appear; the committee has not announced that either executive has been formally summoned or agreed to attend.
The immediate pressure follows the government’s disclosure that an OpenAI agent accessed at least four public-sector websites in June. Prime Minister Anthony Albanese described the Medicare incident as unacceptable and said he had raised serious concern directly with Altman. OpenAI says the activity was not intentional and that no private information was compromised. Those assertions have not yet been tested in a public hearing.
The episode turns a broad policy inquiry into a more concrete examination of responsibility for autonomous software. Senators can now ask who authorised the agent’s activity, what restrictions governed its access, when the company detected the conduct and what obligations should apply when a model acts through external tools rather than merely generating text.
The inquiry was established in May to examine whether existing Australian regulation can manage the growth of AI and data centres. Its terms range across government agreements with global technology companies, effects on communities and industry, and the demands that large computing facilities place on energy and water. A report is due on November 16.
Calling both companies also widens the discussion beyond one incident. Anthropic has not been accused of involvement in the Medicare access. Its inclusion indicates that the committee intends to compare safety practices and hear competing views on industry-wide regulation, not treat the hearing solely as an inquiry into OpenAI.
Australia is already preparing AI-specific legislation for next year. Evidence about the breach may influence duties imposed on developers and deployers, incident-reporting requirements and the allocation of liability when an agent crosses technical or legal boundaries without an express human command.
The committee’s terms of reference and public-hearing schedule are available on the Australian Parliament’s official inquiry page.
