STATEMENTS BY MEMBERS

Ms CHANEY (Curtin) (16:00): We cannot trust OpenAI, Anthropic and other AI companies to regulate themselves, and we can't trust the US government to do it either. This is my clear takeaway from the first week of hearings in the newly established Joint Select Committee on Artificial Intelligence. OpenAI's breaches of Australian government websites and the Hugging Face hack have shown the level of risk and the insufficiency of internal safeguards within these companies.

Australia cannot manage all of the risks, but we can take meaningful steps to improve our chances of making AI a positive force, rather than a dystopian one. For me, three no-brainers came out of the first week of hearings. The first is increasing investment in the Australian AI Safety Institute so we have a credible and useful voice in international standard setting and can futureproof our own regulations.

The second is requiring AI companies to monitor and report key incidents so we know when something goes wrong, rather than waiting months for the companies to decide to tell us. The third is requiring screening of imported nucleic acid sequences to reduce the threat of a terrorist using AI to produce a new virus or pandemic. More will come out of this committee, but the government needs to move quickly to keep up with the escalating risks.

These three actions are no-brainers, Prime Minister; it's time to act.