Australia's top intelligence agency has warned that the country's ageing technology infrastructure leaves it vulnerable to attacks exploiting artificial intelligence, as the government works on new AI regulations. Abigail Bradshaw, director general of the Australian Signals Directorate, told the Sydney Dialogue summit in Canberra that governments and businesses would need to spend "enormous" sums updating outdated systems, including taking offline services people now expect to run continuously, and admitted her agency has no way of knowing how many AI agents are currently active online.

The warning comes amid broader debate over AI safety, with Anthropic chief executive Dario Amodei having called for independent assessors to scrutinise AI models and for US permission for leading firms to coordinate without breaching anti-competition rules. Bradshaw said national security agencies should be involved in any such interrogation. Labor's assistant technology minister, Andrew Charlton, said he took the warnings seriously but resisted letting AI companies or the US alone dictate the rules, with Australia's own guidelines expected early next year; a Greens push for mandatory safety testing laws was rejected by Labor senators the same day. Separately, industry minister Tim Ayres said the government wants more AI training done locally but will not weaken copyright protections, while OpenAI's Ann O'Leary said the company remains keen to deepen ties with Australia.