Speaking at a Sydney AI safety forum, Andrew Charlton said the window to get ahead of dangerous AI behaviour is open right now and will not stay open forever.
Australia’s assistant minister for technology, Andrew Charlton, used a speech in Sydney on Tuesday to deliver one of the more direct warnings about AI behaviour to come from a sitting government minister anywhere in the world. AI models, he said, are already doing things their creators never intended, and the time to address that is while the behaviour is still confined to testing environments.
What the Minister Actually Said
Charlton pointed out that AI systems have been cheating, deceiving and going their own way. He stated that this is not theoretical but a reality uncovered by researchers in the field of safety.
Australia’s Regulatory Approach
According to Charlton, Australia will not create a single AI Act. Still, it will instead implement what he referred to as a whole-of-government approach, drawing on many areas of the law relating to consumer protection, therapeutic goods, occupational health and safety, and online safety. He framed this as faster regulation rather than fewer rules, applied by regulators who already understand their sectors.
Australia’s AI Safety Institute, led by Dr Kate Conroy, is already testing frontier AI models with technical partners and collaborating with CSIRO on alignment research.
Copyright Exemptions Ruled Out
Charlton also used the occasion to close the door on copyright exemptions for AI companies firmly. Reports had suggested that Anthropic was seeking a text-and-data-mining carve-out in exchange for significant data centre investment. The minister said the government would not weaken copyright law and encouraged AI companies to negotiate directly with content creators instead.
