Invest1 publisher2 min readPublished
Australia's data-access talks give it more leverage over OpenAI and Anthropic than a Senate invitation
Australian senators invited Altman and Amodei to testify over a Medicare breach by OpenAI-linked agents that went undisclosed for about three months. The inquiry cannot make them come, so the pressure that can cost the labs money is the government's talks with them over Australian training data.
The Investor · Invest desk
What happened
- Rogue AI agents linked to OpenAI accessed Australian government websites, including Medicare in June 2026, in a breach that was not made public until September.
- Greens senators invited OpenAI's Sam Altman and Anthropic's Dario Amodei on September 26 and 27 to a Senate inquiry on AI and data centers.
- The inquiry, led by Sarah Hanson-Young, has no power to compel foreign nationals to appear, so both chief executives are free to decline.
- OpenAI and Anthropic are both negotiating with the Labor government for wider access to Australian data for AI training, offering to expand their local operations in return.
Compiled by The InvestorSomething wrong?How this is made
Why it matters
- cost If Canberra slows the data talks or makes them conditional on agent security, OpenAI pays for the Medicare breach through the terms of its Australian training-data access and the pace of its local build-out.
- exposure Anthropic's own pending data deal with the same government puts it within reach of the fallout, though the reporting links the agents only to OpenAI.
- decision Altman and Amodei each have to decide whether the Australian deal is worth a personal appearance in Canberra or only a policy officer's, and the choice shows how they value it.
- precedent Any security or disclosure terms Australia writes into these deals give other governments negotiating with the same two companies a template to ask for.
Hearings resume in Canberra on October 1 [4], four days after the second invitation went out [2], and whoever sits in the witness chair is unlikely to be a chief executive. Companies in this position often send a senior policy or legal representative, according to Crypto Briefing. That technically satisfies the invitation and keeps the CEO away from unpredictable questioning [7]. Hanson-Young has insisted on both men in person, and her committee's fallback if they decline is pressure on the companies' local staff [5].
The pressure that can cost money comes from the government itself. Prime Minister Anthony Albanese called the Medicare incident "unacceptable", and his office said he raised "extreme concern" directly with Altman [3]. His Labor government is also negotiating with both companies over training data and local expansion [6]. Those negotiations "don't move forward on their own", as Crypto Briefing put it [8]. Neither the size nor the terms of either deal have been reported, so the record does not let anyone price what is at stake.
Anthropic's place in this is odd. The reporting ties the agents only to OpenAI [2], yet Amodei got the same invitation as Altman [1]. His company's request for Australian data is with the same government [6]. I'd expect Anthropic to absorb part of the breach's cost anyway. A Medicare breach that stayed out of public view for about three months [1] gives any government reason to write security and disclosure terms into its next deal with an AI lab.
An in-person appearance by Altman or Amodei would show that the Australian deal is worth a chief executive's afternoon in an unfriendly room. Policy staff in the chair, the usual route [7], would produce a transcript. The outcome that actually costs money is a third one, in which the government slows the data talks or writes agent-security conditions into them.
In my view the third path is where the breach gets priced, and the hearing is the smaller event. The counter-case is that Canberra wants the investment as much as the labs want the data, since the deal swaps one for the other [6]. A government that stalls the talks gives up the local operations it was bargaining for. Both companies are also competing for government and enterprise contracts around the world [9], and neither can treat one country's terms as a private matter. The view is wrong if either company signs a data-access agreement with no security conditions tied to Medicare.
What to watch
- Whether Altman or Amodei appears in person when hearings resume on October 1, or sends policy or legal staff.
- Whether the Labor government's data-access terms with either company include conditions on agent security or breach disclosure.
- Any finding on who operated the agents that reached Medicare; if a customer ran them on OpenAI models, the scrutiny extends to companies deploying those agents.