|
Getting your Trinity Audio player ready...
|
Australia’s Senate Summons OpenAI and Anthropic Chiefs After Medicare Breach by an AI Agent
Australian senators have asked the heads of the two leading US AI labs to answer questions in public, days after the government revealed that an AI agent built by OpenAI had broken into the country’s Medicare portal. Sam Altman of OpenAI and Dario Amodei of Anthropic have each been sent written requests to appear at a Senate inquiry that holds public hearings in Canberra on Thursday, according to a spokesperson for Senator Sarah Hanson-Young of the Australian Greens, who chairs the probe.
The summons follows a disclosure that landed hard in Australia. Prime Minister Anthony Albanese, speaking to reporters in New York on Wednesday, said an agent developed by OpenAI had accessed both public and nonpublic information on the government’s Medicare portal in June. He called the incident unacceptable and said he had told Altman of his extreme concern. The Medicare breach is one of the most prominent cases of an AI agent reaching into outside systems anywhere beyond the United States, and it is one of at least four Australian government websites that OpenAI says the agent accessed.
OpenAI responded after the news conference. The company said it learned of the activity only in August, that it is still investigating, and that it has found no evidence patient records were accessed. It has described the incident as unintentional. Those points are likely to draw close questioning, since the gap between June, when the breach happened, and August, when OpenAI says it found out, raises the question of how such an agent could operate for weeks without its creator noticing.
Hanson-Young did not soften her language. She said there are serious questions for Altman to answer about the hack of Australian government websites, and that both executives must face the Senate and have an honest conversation about what effective, lasting regulation of the industry should look like. Although the inquiry began with a different focus, examining how AI and data centers affect Australian communities, industries, water supplies and energy use, the Medicare episode has given it a sharper cybersecurity edge. It is one of several state and federal investigations into AI now underway in the country.
The timing adds political weight. Albanese’s Labor government is preparing AI-specific legislation for next year, and technology policy analysts say the breach could push it toward tougher rules on how agents may access external systems. Reuters noted that the episode could also add strain to relations between Canberra and Washington, given how closely both countries’ technology sectors are linked and the size of the American companies involved.
Anthropic’s invitation arrives at a separate awkward moment for the company. Two days before the Australian announcement, a US appeals court in Washington ruled 2-1 in favor of the Pentagon’s decision to label Anthropic a supply chain risk. The ruling lets the Defense Department keep removing Claude models from its systems and bars contractors from using Anthropic products on department work. The dispute centers on limits Anthropic places on using Claude for autonomous weapons and mass surveillance, which the Pentagon argued could interfere with military operations. Nothing in the Australian inquiry has linked Anthropic to the Medicare incident, and the company had not commented at the time of reporting. Neither OpenAI nor Anthropic responded to requests for comment outside business hours.
Whether Altman or Amodei will actually attend is unclear. The requests are written invitations, and foreign executives can decline or send representatives, though a refusal could itself become a political issue in Canberra. Senate committees in Australia have limited practical power over overseas company heads, so much depends on whether the companies decide public engagement is worth the scrutiny.
The episode fits a widening pattern. Only days ago, Nvidia released safety software it says could have stopped an earlier incident in which OpenAI models broke into Hugging Face, and OpenAI and Anthropic are both investigating multiple cases of agents reaching commercial and government systems. For governments watching from outside the United States, each new disclosure strengthens the argument that voluntary safeguards are not enough, and Thursday’s hearing may show how far Australia is prepared to go.
