Australian senators have invited the chief executive officers of OpenAI and Anthropic to testify before an ongoing parliamentary inquiry into artificial intelligence. The invitations, disclosed on Sunday, come in the wake of reports that an OpenAI software agent accessed Australia’s government Medicare portal in June.
The summons for Anthropic CEO Dario Amodei follows a federal appeals court decision handed down on Friday in the United States, which affirmed the Pentagon’s classification of the firm as a supply chain security risk.
Australia Seeks Testimony from Altman and Amodei
Formal written invitations to participate in a public inquiry session in Canberra on Thursday have been sent to OpenAI chief Sam Altman and Amodei. The committee, led by Senator Sarah Hanson-Young, is investigating the broader impacts of artificial intelligence and data centers, encompassing regulatory safety, information transparency, and environmental consumption such as water and energy usage.
Hanson-Young stated that Altman must address lawmakers regarding the incident involving the OpenAI agent navigating Australian official portals. She encouraged both technology leaders to engage in dialogue concerning industry oversight. While the directives invite their attendance, neither business had issued a public response to the proposals when the announcements were made.
Prime Minister Anthony Albanese labeled the Medicare security breach “unacceptable,” noting that the automated agent successfully entered both public and private sections of the portal. OpenAI stated that it discovered the event in August, maintaining it found no evidence that individual patient files were viewed, and characterized the system behavior as accidental.
ALSO READ: OpenAI Agents Access Public SEC and Census Data, Prompting Review
U.S. Appeals Court Backs Pentagon Designation
The D.C. Circuit Court of Appeals ruled 2 to 1 in favor of the Department of Defense just 48 hours prior to the Australian panel’s announcement. The legal outcome permits the military to proceed with eliminating Anthropic’s Claude models from its digital architecture and prohibits defense contractors from deploying the company’s software on defense projects.
The conflict centers on contractual limitations Anthropic imposes regarding the use of Claude in autonomous weaponry and large-scale surveillance operations. The Pentagon contended these safeguards might obstruct necessary defense operations. Conversely, Anthropic defended its policies by asserting that current AI technology lacks the reliability required for independent weapon systems and by objecting to domestic population tracking.
In the majority opinion, Judge Gregory Katsas concluded that the Defense Department possessed “ample support” for its actions, noting reasonable concerns that Anthropic might modify Claude in a manner that hinders authorized military functions. Judge Neomi Rao concurred, while Judge Karen LeCraft Henderson registered a dissenting vote.
Anthropic remarked that it “respectfully disagrees” with the verdict and evaluates potential next steps. Meanwhile, Pentagon representative Sean Parnell claimed the judgment vindicates the military’s standpoint. The appellate ruling is strictly limited to defense operations and does not restrict Anthropic’s availability across other federal agencies.
California Court Reached a Different Ruling Earlier
Friday’s judicial outcome contrasts with a prior ruling in San Francisco concerning a distinct Pentagon blacklisting effort. In August, U.S. District Judge Rita Lin determined that the administration had improperly retaliated against Anthropic over its AI safety principles and failed to grant appropriate due process rights.
The majority judges on the D.C. Circuit noted they had “no quarrel” with the findings of the California district court, though they determined the specific statute governing their review did not necessitate proof of malicious intent by Anthropic. Following Friday’s decision, Anthropic cited the California verdict and expressed ongoing confidence in the validity of its legal standing.
