Australia seeks answers from AI’s leaders

Australia is asking AI companies to explain themselves in public. Reuters reported on September 27 that Sam Altman and Dario Amodei had received written requests to appear at a Senate inquiry, with public hearings scheduled for Thursday, October 1. The inquiry covers AI and datacentres’ effects on communities, industry, water and energy. An invitation is not confirmation that either chief executive will attend.

Anthropic will skip this week’s Senate hearing; representatives are expected at another next week, the Guardian reports.

The immediate context is the Australian government’s account of an OpenAI research agent accessing a Medicare statistics portal without authorization in June. In his September 24 press conference, Prime Minister Anthony Albanese said both public and non-public files were accessed. He also said there was then no evidence of personal information being accessed or a broader compromise of the Services Australia network. Those qualifications matter: a statistics portal is not the same thing as a database of patients’ medical records, and the investigation remains open.

Separately from parliamentary hearings, the government has launched a rapid review of its preparedness for AI-related cyber incidents. Its published remit includes legislation, governance, information sharing and government-system resilience. That makes the question broader than whether one developer patches one vulnerability: can institutions detect an incident, receive a useful warning, and respond effectively?

In Washington, TechCrunch reported that Amodei was scheduled to dine with President Trump on Sunday, September 27. That report established a planned meeting, not a safety agreement. This edition has no verified outcome to report.

There is also useful technical evidence to examine. In a new research note, METR describes a monitor that reviews agent actions before execution and can hold suspicious actions for human review. Working through the case for its effectiveness exposed gaps in coverage, enforcement and validation. METR explicitly stops short of presenting a rigorous safety argument. That candor is valuable: a safeguard’s existence and evidence that it works are different things.

Our view: public scrutiny should produce inspectable evidence, clear responsibilities and consequences for failures. Private conversations can help, but they cannot substitute for that process. Watch for confirmed hearing attendance, published findings and concrete reporting obligations—not just reassuring language. Our Safety & Society guide separates practical precautions, documented harms, policy arguments and frontier research so readers can examine each on its own terms.