Background
In May 2026, the United States Studies Centre (USSC), in partnership with the Department of Foreign Affairs and Trade and the Department of Industry, Science and Resources, hosted a Track 1.5 Dialogue on Artificial Intelligence (AI) in Canberra.
The USSC first convened a Track 1.5 AI Dialogue in 2025.1 Since then, the global AI landscape has evolved significantly. AI capabilities have continued to advance rapidly, governments are releasing AI strategies, and competition over compute, infrastructure and supply chains has intensified. Australia has also released its National AI Plan2 in December 2025 and committed to developing a strategy for international engagement and regional leadership on AI. As AI becomes increasingly central to economic competitiveness, national security and geopolitical influence, questions of capability, resilience and regional AI diplomacy have become more urgent.
Against this backdrop of accelerated progress and intensifying competition, USSC partnered with the Australian Government to bring together 40 experts from government, industry and academia to consider Australia’s domestic and regional role in shaping the future of AI. Discussions examined opportunities and challenges across the AI stack, including infrastructure, compute, talent and supply chains, as well as practical pathways for regional cooperation across the Indo-Pacific. The following report highlights the key themes and policy insights that emerged from the discussions.
Supported by
The views expressed herein are not necessarily the views of the Commonwealth of Australia, and the Commonwealth of Australia does not accept responsibility for any information or advice contained herein.
Theme 1: AI Capabilities continue to accelerate
AI capabilities are advancing at an unprecedented rate across a growing range of domains. Existing benchmarks are being surpassed, with advances in model performance continuing to outstrip traditional measures of progress (see Figure 1).3 Today’s AI systems are the least capable they will ever be, underscoring both the pace of technological change and the difficulty of forecasting future developments. Indeed, participants noted the prospect of artificial general intelligence (AGI) — broadly understood as an AI system capable of learning, reasoning and applying knowledge across domains at a level comparable to or exceeding human performance — is already a serious objective among frontier AI companies and policymakers in the United States.
As capabilities improve, the range of tasks that AI systems can perform continues to expand. While this creates significant opportunities in scientific discovery, medical research, public service delivery and productivity gains, it also introduces new risks. Many advanced AI capabilities are inherently dual-use. Advances in cybersecurity, for example, can strengthen defensive systems while simultaneously enabling more sophisticated offensive attacks. Similar concerns apply across domains, including military applications, supply chain management and the potential misuse of AI-enabled biological tools.
Anthropic’s Claude Mythos Preview illustrates both the pace and unpredictability of AI development.4 The model demonstrated a significant advancement in cyber capabilities, with the ability to identify software vulnerabilities and to develop increasingly sophisticated methods to exploit them.5 Importantly, these capabilities were not deliberately trained but emerged as a by-product of training for general reasoning and coding tasks, highlighting how novel capabilities may emerge unexpectedly as frontier models continue to improve. Mythos Preview also highlighted the need to gatekeep access to these powerful capabilities. Project Glasswing,6 Anthropic’s collaborative initiative involving major technology and cybersecurity firms, was established to apply Mythos to the defence of critical software systems and digital infrastructure while keeping the model out of the hands of cyber attackers.
A key implication for governments is the need to prepare for a range of possible AI futures. Participants highlighted that the non-linear pace of AI progress, combined with the emergence of novel, unanticipated capabilities, creates uncertainty around the timing, nature and scale of future economic, social and national security impacts. Government responses will therefore need to be adaptive, with policy, regulatory and investment frameworks capable of responding to increasingly unpredictable technological developments.
Theme 2: Competition is heating up over global AI supply chains
The development and deployment of AI are shaped by access to a complex ecosystem of inputs, including advanced semiconductors, data centres, energy systems, cloud infrastructure and specialised talent. Competition over these enabling resources has intensified. Today, the United States and China sit at the frontier of AI development (see Figure 2) because they have unique advantages and points of leverage across key components of the AI ecosystem. This concentration of capability is increasingly translating into geopolitical influence, as states seek to secure advantage across the AI value chain and shape the rules governing its development.
The United States continues to lead in frontier model development and advanced semiconductor design, while also leveraging export controls and strategic partnerships to maintain its advantage.7 China, meanwhile, is releasing increasingly competitive models and pursuing a strategy of technological self-reliance while expanding its global influence through infrastructure investment, technology exports and participation in international standards-setting processes. Initiatives like China’s Digital Silk Road, a component of the Belt and Road Initiative that seeks to expand China’s technological reach overseas, demonstrate how digital infrastructure is increasingly being used as a tool of economic and strategic influence, particularly across developing countries.8
At the same time, the interconnected nature of AI supply chains limits the feasibility of complete technological self-sufficiency. Developing advanced AI systems depends on a globally distributed supply chain where no single country controls every link in the chain. Semiconductor supply chains illustrate this challenge, with critical capabilities concentrated across multiple jurisdictions and production processes that are highly specialised and interdependent (see Figure 3).9
In this environment, complete AI sovereignty is neither realistic nor necessarily desirable. Rather than pursuing full ownership of the AI stack, AI ‘agency’ was emphasised as a more practical objective. For countries such as Australia, the objective of AI agency is to maintain reliable access to critical capabilities, preserve choice between suppliers, ensure that essential systems cannot be easily disrupted, and develop the leverage needed to shape and capture value from AI development. This shifts the focus from self-sufficiency to resilience, diversification and trusted partnerships, making AI agency both an economic and national security imperative.
Theme 3: Australia has opportunities to build AI agency, but must move fast
While Australia is unlikely to compete directly with the United States or China across every layer of the AI stack, it possesses a range of assets that provide opportunities to shape outcomes, capture value and strengthen resilience. Replicating the capabilities of the United States or China would be unrealistic. However, Australia can build influence and become an indispensable participant within key parts of the AI ecosystem.
Australia's strengths span several layers of the AI stack.10 These include world-class research institutions, valuable data assets, abundant renewable energy resources, growing data centre infrastructure, critical mineral reserves and trusted international partnerships. Australia also possesses expertise in areas such as robotics, computer vision, cybersecurity and defence-related applications.11 Together, these strengths provide a foundation for building AI agency.
One of Australia’s most significant opportunities lies in data centre and compute infrastructure. Compute, referring to the processing power required to train and operate large-scale AI models, lies at the centre of the AI ecosystem. As AI models become larger and more resource-intensive, access to compute is emerging as both a key determinant of technological leadership and a bottleneck. At the same time, data centre expansion in several overseas markets, including the United States, is facing constraints related to energy availability, permitting processes and growing community opposition. These pressures create opportunities for countries that can offer reliable energy, political stability and favourable investment conditions. Australia's renewable energy potential, available land and strategic location position it well to attract investment in AI infrastructure and become a regional hub for AI data centres.
However, realising this opportunity will require addressing a number of domestic constraints. Regulations around copyright and data use, for example, were identified as a significant barrier to AI training and investment. Australia has ruled out a text and data mining exception for AI,12 and current copyright laws create legal uncertainty for AI training in Australia, exposing firms to potential lawsuits over copyrighted material that deters investment and drives development activity elsewhere. Additional challenges include workforce capability gaps, bureaucratic bottlenecks and difficulties in commercialising research. Over the past 20 years, approximately 11,000 startups and other businesses have left Australia for the United States, United Kingdom and Canada, often to seek out better market opportunities and access to capital.13
Public trust also remains an important consideration. Only 36% of Australians trust AI systems, a figure 10% below the global average and lower than other countries actively pursuing AI leadership agendas, including China, the United States and the Republic of Korea.14This comparatively low level of confidence in AI may potentially limit adoption and reduce the benefits that could otherwise be realised through technological investments.
The importance of speed emerged as a recurring theme, especially as competition for AI investment is intensifying and countries are actively positioning themselves as destinations for AI infrastructure and compute. Canada has expanded incentives for the AI value chain and data centre infrastructure through its National AI Strategy;15 Japan plans to significantly increase public investment in domestic semiconductor and AI development,16 and the Nordics have leveraged their access to cost-efficient renewable energy and cool climates to attract capital for hyperscalers, the large data centres optimised for AI.17 As other governments increasingly compete to attract AI-related capital and infrastructure, delays in policy reform and infrastructure development risk reducing Australia’s ability to capture value from the rise of AI. Australia may find itself falling further behind as peer competitors pull ahead. Building AI agency is therefore not a long-term ambition, but an immediate strategic priority requiring coordinated action.
Theme 4: Regional AI leadership will depend on trusted partnerships and locally driven AI development
Australia's role in the regional AI ecosystem will be shaped not only by the strength of its domestic capabilities but also by its ability to engage constructively with partners across the Indo-Pacific. The Dialogue emphasised that regional AI leadership is an important component of Australia's economic, technological and strategic positioning. As AI becomes more deeply embedded in infrastructure, supply chains and governance frameworks, the ability to shape regional adoption pathways through partnerships, standards and infrastructure will become an increasingly important source of influence.
Across Southeast Asia and the Pacific, interest in AI is growing rapidly, driven by the potential to improve productivity, strengthen government services, enhance disaster response and support economic development. However, levels of digital maturity, infrastructure readiness and institutional capability vary significantly across the region (see Figure 4). These differences mean that regional engagement strategies must be tailored to local contexts rather than applying a uniform approach.
In Southeast Asia, strong interest in AI is being driven by rapid digital growth, increasing private sector investment, an ambitious start-up sector and the proliferation of national AI strategies. Six of the eleven member states of the Association of Southeast Asian Nations (ASEAN) have established national AI strategies18, reflecting growing recognition of AI as a driver of economic competitiveness and digital transformation. Vietnam’s AI Law19, released in March 2026, introduced the region’s first legally binding framework on AI, which seeks to balance innovation with safety and risk management.20 While countries are exploring how AI can improve productivity, public services and economic competitiveness, many continue to face constraints in infrastructure, workforce capability, regulatory implementation and access to finance. In this context, opportunities for cooperation should likely centre on technology partnerships, capability uplift, digital infrastructure and governance frameworks that support responsible AI adoption.
The Pacific, however, presents a very different set of priorities. While Fiji has a National Digital Strategy and Papua New Guinea is developing an AI adoption framework, most Pacific Island states have not developed formal AI strategies or regulatory frameworks for safe AI deployment.21 Many Pacific Island countries continue to face foundational challenges relating to connectivity, energy reliability, digital infrastructure, data availability, cybersecurity and institutional capability. As a result, discussions about AI are often inseparable from broader questions of basic digital development. The Dialogue highlighted that AI infrastructure in the Pacific must be understood holistically, encompassing not only data centres and compute capacity, but also resilient connectivity, trusted data management, governance systems and skills.
A recurring theme was the importance of moving beyond fragmented pilot projects towards more coordinated and long-term approaches to digital and AI development. Pacific partners, in particular, expressed a preference for investments that address underlying structural constraints rather than isolated demonstrations of technology capability. In many cases, investments in connectivity, power infrastructure, cybersecurity and digital literacy may deliver greater near-term impact than direct investment in advanced AI systems.
Trust, sovereignty and local ownership were also identified as critical enablers of AI adoption. Participants noted that data sovereignty, particularly in the Pacific, is closely linked to governance, cultural values and control over sensitive or community-owned information. Ensuring that AI systems reflect local priorities and operate within trusted governance frameworks will be essential to building confidence and supporting long-term adoption. This includes protecting cultural heritage, respecting local decision-making processes and ensuring that technological development strengthens, rather than undermines, local agency.
Regional engagement must also be understood within the context of growing strategic competition. Many countries across Southeast Asia and the Pacific are seeking to balance relationships with multiple partners rather than align exclusively with any one country. Australia has opportunities to position itself as a trusted and reliable partner. Australia's comparative advantage lies in listening to regional priorities, supporting locally driven development and delivering holistic outcomes through trusted partnerships with other donors and trusted firms. Participants also commented that Australia should avoid any perception that it is forcing countries to make binary choices in the technologies they adopt.
Australia's experience in areas such as cybersecurity, online safety, digital regulation and technology governance provides a strong foundation for regional engagement. However, leadership will ultimately depend less on exporting Australian solutions or directing outcomes and more on enabling partners to pursue their own locally-defined objectives.
Australia’s comparative advantage lies in listening to regional priorities, supporting locally driven development and delivering holistic outcomes through trusted partnerships with other donors and trusted firms.
