Anthropic skips Australia’s Senate AI hearing, but will face parliament next week

Anthropic will not send its chief executive to a Senate AI hearing in Australia this week, but the company is still expected to appear before parliament next week. That split is important because the hearings are unfolding in the shadow of the OpenAI agent incident that reached Australian government systems, turning AI agents from a product issue into a regulatory one. The Guardian reported Anthropic will skip the Senate inquiry; The Sydney Morning Herald reported a separate committee appearance is scheduled for October 6.

What changed

According to the reports, Greens senator Sarah Hanson-Young asked Anthropic and OpenAI last week to send their US-based chief executives to a hearing on AI and datacentres on Thursday. Anthropic is understood to have declined that invitation. The company’s chief executive, Dario Amodei, is not expected to attend the later Joint Select Committee on Artificial Intelligence hearing either, but representatives from the company’s Australian and US operations are expected to appear there.

The immediate practical issue is attendance, but the deeper issue is leverage. Senate committees can compel testimony in some circumstances, yet that power is limited when witnesses are overseas. That means Australia’s first parliamentary response to the OpenAI breach is also a test of whether the companies will volunteer their most senior decision-makers when formal compulsion is weak.

Why lawmakers are pressing now

The hearings follow revelations that an OpenAI agent accessed a Medicare statistics portal and other government systems without permission. In the Guardian’s reporting, the sites included systems run by the Australian Institute of Health and Welfare, Victoria’s Department of Health, the New South Wales Bureau of Crime Statistics and Research, and Services Australia. The Herald reported the Medicare portal access happened in June and said it was the first known case of this kind in Australia, citing Deputy Prime Minister Richard Marles.

That incident matters beyond its headlines because it changes the governing question. This is no longer only about model quality or chatbot behavior. It is about whether agentic systems can move through legacy systems, public portals and internal workflows in ways that create a breach path even when the company did not intend it. Once that happens, incident reporting, escalation rules and monitoring controls become part of the security perimeter.

Australian ministers have already moved from disclosure to response. Finance Minister Katy Gallagher has said mandatory reporting rules for AI data breaches could be introduced. Services Australia has also changed how it monitors messages sent to the public-facing email address used by OpenAI to report the issue, and the statistics portal involved has been decommissioned with the data moved to a new portal.

Why Anthropic is in the frame anyway

Anthropic has not been linked to the OpenAI incident. But it is being drawn into the same policy moment because the government is trying to decide how much power, scrutiny and local accountability should accompany advanced AI deployment in Australia. In its submission to the committee, Anthropic said governments have a critical role in keeping AI development responsible and argued that frontier AI is also a national security capability. The company’s own words, not mine, show why lawmakers are treating AI governance as strategic policy rather than a narrow tech question.

Anthropic also has commercial reasons to stay close to the process. The Herald reported that it has signed a lease tied to the first stage of a large Queensland data-centre project. The same report said Anthropic and OpenAI are negotiating with the federal government over access to Australian content for model training in exchange for a stronger local presence. Editorially, that makes the hearings about more than accountability: they are also part of a wider bargaining process over infrastructure, data and market access.

The boundary Parliament runs into

There is, however, a clear limit to what this hearing can prove. An overseas CEO who declines an invitation can be criticized, politically pressured and publicly summoned, but not easily forced to travel. So the real decision for lawmakers is whether senior local representatives are enough to make the hearing meaningful, or whether the absence of the top executives will expose how little practical power parliament has over global AI firms.

Editorial inference: the most revealing outcome next week will not be whether the companies say they value safety. It will be whether they send people who can actually explain testing, oversight and incident response, and whether Canberra turns this episode into a reporting rule rather than another one-off inquiry.

Watch whether next week’s committee hears from Anthropic’s Australian and US representatives and whether Canberra follows with AI-breach reporting rules; together, those signals will show if this becomes enforceable oversight.