in briefThe Australian Signals Directorate has revealed that it has access to high-level, non-public AI models.Australia should use its privilege and power to help others defend against AI attacks, the country's top cyber spy chief says.Australia's top cyber spy chief has said the country has a "technical edge" in dealing with the risks posed by frontier artificial intelligence models and should be training other nations to defend against them.Speaking after a recent apology by United States technology firm OpenAI for a breach of Australian Medicare data by one of its experimental AI models, Australian Signals Directorate (ASD) chief Abigail Bradshaw said Australia's Five Eyes membership has resulted in key advantages from powerful US tech companies. The director-general of the ASD, which manages the country's national cyber security and foreign intelligence gathering, told ABC Radio National on Wednesday that her team had been granted access to developmental AI models over six months ago. "Australia has enjoyed a rare level of engagement with the frontier labs," she said. "We have had an incredible alliance starting with the Five Eyes Intelligence Alliance."Few other nations have been granted access to high-level, non-publicly available AI models, which Bradshaw said her team has been using to shore up government systems ahead of the rest of the world. News that makes senseYour trusted source for staying up-to-date with the world around you. Get free daily news updates and analysis, straight to your inbox.Enter email addressSubscribeMorning (Mon–Fri)Afternoon (Mon–Fri)WeekendInsightBy subscribing, you agree to SBS’s terms of service and privacy policy including receiving email updates from SBS.Now, the country should be sharing that information with others, she stated.Five Eyes is an intelligence-sharing alliance of Australia, New Zealand, Canada, the United Kingdom and the US, which developed after World War Two. In June, the alliance released a three-page document saying the pace of AI development means major operational shifts are "months away, not years" and that organisations should "assume breaches will occur". Prime Minister Anthony Albanese told the United Nations last week that the world needs greater cooperation in dealing with the threats of AI. Source: AAP / Mick Tsikas"For the last 80 years, we have enjoyed a technical edge as a consequence of our membership and our partnership in the Five Eyes alliance," Bradshaw said. She noted, however, that "a new partnership" is required between governments, research groups, industry, and frontier AI companies like OpenAI, Anthropic, and Google DeepMind due to the nature of the threat. The ASD Cyber Security Partnership Program, which aims to address the issue, was reported as having 133,000 members last year, but Bradshaw said this has now grown to 150,000 as organisations come together to share information about AI.OpenAI apology 'not insignificant'Last week, Prime Minister Anthony Albanese announced at the United Nations General Assembly that an OpenAI model had breached a Medicare site.The incident took place in June, but the company didn't realise until August and failed to notify the government until 10 September. The company issued a formal apology on Tuesday, pledging technical support and the establishment of a local task force to advise on AI safety. Its newest frontier model, GPT-6.1 Astra, was shelved over safety concerns.In response, the government has launched a rapid review of the incident and issued a directive to all federal departments to conduct an immediate stocktake of legacy IT systems. A review of existing security reporting laws is also taking place.Bradshaw said the apology and cooperation of OpenAI is encouraging. "Look at the content of what OpenAI have set out there; that is not insignificant," Bradshaw said."In fact, that is consistent with our experience, with our engagement with not just OpenAI, but the other labs, Anthropic and others, over the last six months. "That's a really important precedent for the way this relationship proceeds into the future."Engagement with frontier AI companies is a "two-way street", Bradshaw said, but added that the technology could also provide advantages if those relationships are managed well. "Cyber security is like painting the Sydney Harbour Bridge," she said. "It never stops."For the latest from SBS News, download our app and subscribe to our newsletter.Share