Under the agreement, both countries will share research and develop joint approaches to evaluating AI capabilities and potential cyber-security threats. <\/p>\n
The United Kingdom and Australia are deepening cooperation<\/a> on AI security through a new partnership between the UK AI Security Institute and the Australian AI Safety Institute.<\/p>\n Under a Memorandum of Understanding, the two institutes will share information on frontier AI capabilities, collaborate on AI evaluation practices and exchange research findings. The UK government said the partnership will focus partly on how advanced AI systems could be used in cyberattacks, as well as how they can strengthen defensive capabilities.<\/p>\n The agreement will also support staff exchanges between the two institutes, strengthening day-to-day collaboration. UK officials said the partnership reflects the need for trusted international cooperation as AI systems evolve quickly and create new security and safety risks<\/a>.<\/p>\n The UK\u2019s AI Minister Kanishka Narayan is expected to sign the agreement with Australia\u2019s Assistant Minister for Science, Technology and the Digital Economy, Andrew Charlton, during a meeting in Canberra. Narayan said no country can address fast-moving AI risks alone, particularly in cybersecurity.<\/p>\n The announcement follows research from the UK AI Security Institute showing that advanced AI systems are rapidly improving their ability to carry out complex cyberattacks, creating opportunities for both attackers and defenders. The UK said the institute\u2019s frontier AI research continues to inform policymaking to protect businesses, critical infrastructure, and the public.<\/p>\n Why does it matter?<\/p>\n The partnership shows how AI security is becoming a matter of international coordination, especially as frontier models develop stronger cyber capabilities. By sharing research, evaluation methods and staff expertise, the UK<\/a> and Australia<\/a> are trying to reduce blind spots in oversight and develop more consistent approaches to testing fast-moving AI systems.<\/p>\n