During a parliamentary inquiry on Friday, Liberal MP Andrew Hastie shared an incident involving Microsoft Copilot to illustrate the risks associated with foreign-run artificial intelligence. Hastie reported that when he used the AI tool to draft a response to a constituent who had written to him about their intention to end their life through voluntary assisted dying, the software suggested inappropriate replies. Specifically, the AI proposed responses such as “congratulations!”, “great to hear from you,” or “that is wonderful news!”

Hastie presented this example to emphasize the potential dangers of relying on American-run software for sensitive political and personal communications. He argued that these errors highlight the shortcomings of current AI models and used the experience to advocate for the development of Australian-run AI infrastructure. The inquiry is currently examining the Albanese government’s efforts to attract major AI firms, such as OpenAI and Anthropic, to establish a stronger presence in Australia. Hastie’s testimony serves as a cautionary note regarding the integration of these technologies into the daily operations of government offices, suggesting that current safeguards may be insufficient for handling sensitive constituent correspondence.