The recent AI sovereignty debates sparked by the Trump administration's move to restrict access to advanced AI models have left many nations, including Australia, pondering their AI strategies. In this article, I'll delve into why Australia should focus on influencing the sharing of AI models rather than attempting to build its own frontier AI.
The Challenge of Frontier AI
Frontier AI, with its cutting-edge performance, is a game-changer, especially in cybersecurity. The recent export controls on Anthropic's Mythos and Fable models highlight the potential risks and the scramble for access.
Australia, as a nation heavily reliant on American technologies, faces a unique challenge. The prospect of an informal and opaque licensing regime for AI models is a cause for concern, as it could lead to delayed access or even sudden revocations.
Why Australia Shouldn't Go It Alone
The costs of developing frontier AI are astronomical, and Australia lacks the natural advantages that other countries possess. Investing in frontier AI would be an inefficient use of resources, especially when considering the experience of countries like France and the UAE, which have invested billions but haven't yet matched the performance of American or Chinese models.
Performance matters, especially in national security applications. Accessing the best models quickly is crucial for defense and deterrence.
A Pragmatic Approach
Instead of competing at the frontier, Australia should focus on areas where it has more maturity and agency. Investments in critical minerals, data center infrastructure, and specialized datasets for AI training are more feasible and beneficial.
Shaping Access
Australia should advocate for trusted allies to have priority access to powerful models. The UK's reported lobbying for access to future models and the G7 Summit discussions are steps in the right direction.
A tiered access system could be an effective solution, ensuring that the most powerful models are used for defensive purposes by vetted organizations. Independent experts can evaluate models and recommend safeguards, creating a transparent system.
Australia's AI Safety Institute can play a crucial role in this process, contributing to the technical evaluation and recommendation of safeguards.
The Bigger Picture
Frontier AI provides an opportunity for allies to contribute to collective security. Hoarding these technologies could undermine the reliability of the United States as a technology supplier and security partner.
Australia must act now to shape the rules of access. The next frontier AI model is on the horizon, and the country must decide whether to be proactive in influencing these rules or risk being left behind.
