One of Australia’s leading intelligence agencies has raised concerns about the risks posed by artificial intelligence (AI) to the country’s outdated technology infrastructure. Abigail Bradshaw, director general of the Australian Signals Directorate, warned at the Sydney Dialogue summit in Canberra that the number of AI systems operating online is largely unknown. “I think the idea of getting a clear count or understanding of how many of these AI agents are out there is really remote,” she said. “And I would just say: a lot.” Bradshaw emphasized that Australian governments and businesses may need to invest heavily in updating their systems, some of which are currently vulnerable and could be exploited by AI threats. At the same time, Anthropic’s chief executive, Dario Amodei, has called for independent third-party assessments of AI models and for the US government to allow major AI companies to coordinate on safety standards without facing legal penalties for anti-competitive behavior. Bradshaw suggested that national security agencies should also be involved in reviewing AI systems for potential risks. Meanwhile, the Greens pushed for mandatory safety testing and transparency laws for advanced AI systems, but their proposal was rejected by Labor senators. Andrew Charlton, Australia’s assistant minister for technology, acknowledged the concerns raised by AI companies but said the government should not rely on US or corporate leaders to set AI regulations. “I don’t think we should care very much about what individual chief executives say,” Charlton said. He added that the Albanese government plans to introduce AI guidelines early next year, covering both the risks of extremely dangerous AI systems and more routine applications. Anthropic’s chief lawyer, Jeff Bleich, expressed optimism about Australia’s role in shaping global AI rules, arguing that decisions on AI should not be left to just a few companies or countries. He said that while creating such rules would be difficult, it was “better to come to Australia.” Another issue influencing AI companies’ interest in Australia is the use of locally created content—protected by copyright—to train AI models. Industry Minister Tim Ayres said the government supports training AI models using local data but does not want to weaken copyright protections. OpenAI’s global policy vice-president, Ann O’Leary, said her company is engaged in ongoing discussions with the Australian government about these rules. She emphasized the importance of diversifying where AI models are trained and acknowledged Australia’s strategic significance. Meanwhile, Coalition leader Andrew Hastie warned that without attracting advanced AI companies to Australia, the country could become dependent on foreign technology in future conflicts. He called for a formal partnership with the US, possibly through the Aukus security alliance, to bring major AI companies to Australia. However, a defense expert suggested that without developing its own cutting-edge AI models, Australia might remain a dependent rather than an independent technological power.