CSIRO emphasizes the need for APAC Region to establish its own AI foundation models

Research conducted by the Large European AI Models initiative in 2023 revealed that the majority of AI foundation models since 2017 were from the U.S., while China accounted for a smaller percentage. It emphasized that Germany’s lack of an AI “paradigm shift” could potentially threaten Europe’s digital sovereignty. This dominance of AI foundation models by the U.S. and China is also raising concerns in APAC countries, with worries about the impact on data security, governance, and geopolitical and supply chain risk. As a response, countries like Singapore, Japan, and Australia are exploring the development of their own sovereign AI foundation models. This move could benefit businesses by providing models more attuned to their culture and language while keeping data secure within national borders. The U.S. continued to outpace China in the development of AI models in 2023, with Stanford’s AI Index Report finding a significant number of notable models originating from the U.S. Australia, Singapore, and Japan are among the countries in APAC that are creating or considering sovereign AI models. The concentration of AI foundation models poses risks for businesses in APAC, including reliance on dominant providers and exposure to unexpected changes. There is also a growing recognition of the importance of cultural relevance, sensitivity, confidentiality, privacy, security, geopolitical risks, and ethics in AI models. While sovereign AI poses risks, it also presents opportunities for businesses. APAC countries, such as Singapore, Japan, and Australia, are taking steps to rebalance the global foundation model market by developing their own bespoke AI models. For instance, Singapore has initiated a program to develop a regional AI model, while Japan is partnering with firms to create AI models tailored to the Japanese language and culture. Australia is considering various measures to strengthen its position in the global AI order, including investing in computing infrastructure, datasets, AI skills training, and international collaborations.