The Pulse
Australian Inquiry Exposes Microsoft Copilot’s Dangerous Advice
Australian MP Andrew Hastie told a parliamentary inquiry that Microsoft Copilot suggested replying “congratulations!” to a constituent planning voluntary assisted dying. The hearing also examined Australia’s dependence on US-based AI system
AI.info Team ·
Copilot’s suggested reply puts AI judgment under scrutiny
Microsoft Copilot suggested that Australian MP Andrew Hastie respond “congratulations!”, “great to hear from you” or “that is wonderful news!” after a constituent with a terminal illness wrote that he planned to use voluntary assisted dying.
Hastie raised the exchange on Friday, September 18, during Australia’s parliamentary inquiry into artificial intelligence. He used the example to argue that systems operated by American companies can fail badly when they encounter sensitive personal and cultural situations. The Guardian reported the exchange from the hearing.
The episode does not establish that Copilot caused harm or that the constituent acted on its advice. It does show the danger of treating a general-purpose writing assistant as a reliable adviser for messages involving death, illness or other vulnerable circumstances.
Andrew Hastie argues for Australian control
Hastie used the incident to support his call for AI systems that are developed and operated in Australia rather than relying entirely on services hosted and controlled overseas. The Liberal MP has also argued that Australia should secure access to advanced models while requiring companies to operate under Australian law.
Australia’s Joint Select Committee on Artificial Intelligence was appointed on August 20, 2026. Its terms include sovereign AI capability, the adoption of AI by government agencies, data sovereignty, the effects of AI on vulnerable people, and national-security risks. The committee is due to report on November 30, 2026, according to the Parliament of Australia’s committee page.
The dispute at the hearing was not simply about whether AI should be used. Government and defence officials argued that Australia also faces risks if it cannot access the most capable systems.
Cyber officials warn against cutting off advanced models
Abigail Bradshaw, director general of the Australian Signals Directorate, told the inquiry that her agency uses advanced AI to defend against what she called “highly capable malicious cyber actors.” She said the agency is entirely reliant on processing housed in the United States and could lose substantial capability if access were cut during a conflict.
“If a digital adversary has access to newer, more capable models than we do, they may be able to identify and exploit vulnerabilities in systems and networks before we can identify them,” Bradshaw said, according to the Guardian’s account of the hearing.
Her warning set up the central tension in the inquiry. Australia wants greater control over data, computing infrastructure and model access, but officials also say that excluding leading US developers could leave Australian agencies with weaker tools than the actors they are trying to stop.
Defence wants compute, data and decisions kept in Australia
Chris Crozier, the defence department’s chief AI officer, said the government was pursuing a domestic infrastructure plan involving Google and a network of linked data centres. The facilities would be separated from the United States, with the aim of ensuring that Australia’s systems could not be switched off by a decision made overseas.
“We do not want to be using AI tool sets that are located offshore … we want the compute, we want the application, we want the data, we want the decision making, all based in Australia.”
Chris Crozier, chief AI officer, Australian Department of Defence, as quoted by The Guardian
Officials told the committee that bringing major AI companies into Australia could also make it easier for law-enforcement agencies to respond to abuse and other threats on their platforms. The government is consulting on standards that could require companies to report safety incidents and breaches, reserve computing capacity for Australian research, and meet conditions tied to the construction of large data centres.
Copyright rules add another fight
The hearing also exposed a dispute over the data used to train AI systems. Jenna Priestly, an assistant secretary at Australia’s attorney general’s department, said proposed copyright changes could affect access to foreign-made material distributed in Australia as well as Australian works.
Priestly said developers wanted to train models on “maybe billions” of pieces of information and argued that negotiating with every rights holder would be practically impossible. Creative groups have opposed any default system that would allow companies to collect Australians’ online work without licensing agreements.
The government has said it wants draft legislation covering broader AI standards by the end of 2026. Bradshaw told the inquiry that decisions about training and access were urgent because developers could choose other countries if Australia delays.
Hastie’s Copilot example gives that policy debate a human scale: a system designed to help draft an ordinary message produced advice that was plainly unsuitable for a constituent discussing the end of his life. The inquiry now has to weigh that failure against the security argument that Australian agencies cannot afford to fall behind the systems used by hostile cyber operators.