By Allyson Ye and Rorry Daniels
Hi there. In today’s Asia Policy Brief, ASPI’s Summer Endowment Intern Allyson Ye and Managing Director Rorry Daniels examine President Trump’s U.S. AI Action Plan and the potential ramifications for Southeast Asian countries. While pursuit of ‘AI sovereignty’ presents its challenges, regional frameworks can provide unique opportunities for Southeast Asia’s technological agency amidst U.S.-China competition.
State of Affairs: Trump’s Zero-Sum AI Vision
“America started the AI race. And we’re going to win it.”
So heralded Donald Trump on July 23 when he signed the U.S. AI Action Plan into being, signaling the administration’s embrace of a competitive, deregulatory approach to artificial intelligence (AI). Underpinned by three executive orders, the action plan strips red tape such as environmental review, removes “woke AI” from government, and expedites the buildout of large-scale domestic AI infrastructure. Of most interest to international players, the plan calls for the U.S. to develop and promote its own open-source AI models—an area where China has historically made inroads with countries wary of ‘black box’ U.S. tech—and also reverses the Biden administration’s restrictions on advanced-capability AI, calling instead for hardware, software, and U.S. standards to be bundled into a “full-stack export package” for worldwide adoption. Altogether, the plan centers on achieving U.S. supremacy over the global AI market. The pernicious consequences if it fails? China will win the AI race and dominate the 21st century global economy.
In Southeast Asia, the volatility of U.S.-China geopolitical jockeying has made ‘AI sovereignty’ a key strategic priority, as regional states seek to detangle themselves from a binary AI race by chasing geopolitical independence in domestic AI stacks. However, Trump’s action plan seems to take for granted that third countries are helpless to accept the rules of the game—that the future of AI geopolitics will be written by bipolar superpower rivalry, and the only question is with whom will they side. If the global market casts smaller countries overwhelmingly as buyers rather than partners on the AI value chain, Southeast Asia may find ambitions of sovereign AI increasingly costly.
Why It Matters: The Cost of Alignment
Invocations of ‘AI sovereignty’ made in the promotion of Indonesia’s Sahabat-tailored large language model, Thailand’s SIAM.AI Cloud, and Malaysia’s first full-stack sovereign AI infrastructure, are all driven by recognition of the glaring security risks of ceding too much of this critical technology to foreign control. In an age where compute is revered as a critical resource akin to oil, overdependency on a foreign AI stack is naturally an alarming prospect. What happens if the service is cut off? What happens if the dependency is exploited by the provider country? A domestic market hyper-attached to one foreign vendor leaves smaller countries uniquely vulnerable to not only price fluctuations and supply shocks, but also political leverage.
Perhaps more tangibly felt are the implications of AI sovereignty for public trust and AI adoption. If AI is the ‘next big thing’ characterizing frontier nations, public uptake of AI is crucial for smaller countries to evolve alongside the global economy. However, AI products imported wholesale from abroad can contain baked-in biases unique to the developer’s country that impede their own adoption. Common commercial models are trained on predominantly Western datasets, and aside from posing conversational barriers to non-English users, have been shown to reinforce sweeping generalizations and racial stereotypes about Asia learned from their Western cultural corpus. AI standards imposed by the Global North may not map onto how concepts like ‘privacy’ and ‘explainability’ are construed in collectivist cultures. These barriers to AI uptake could easily worsen the global digital divide, as Southeast Asian countries struggle to integrate AI amenable to their language, culture, and workforce.
If smaller countries do scale back national ambitions and tie themselves into one of two AI ecosystems top to bottom, this bifurcated tech landscape might affect the diffusion of ideas. Global division into two AI markets of limited interoperability might create diverging sets of information, content, and applications available to users, a two-party ‘echo chamber’ effect all the more troubling considering AI’s integration into media and education, and its potential for ideological weaponization. In turn, refining the models based on training data limited to each ecosystem may entrench biases that drive the two further apart.
What to Watch: A Non-Sovereign Solution?
Even absent great power competition, ambitions of sovereign AI are hampered by a swathe of practical concerns. AI infrastructure is financially costly, with worldwide capital outlays forecasted to reach almost $7 trillion to meet demand for compute. Data centers guzzle frightening amounts of water and display a voracious appetite for energy that has countries hankering to develop next-generation sources, which themselves require years to build. Compounding these limitations with the increasing pressure to align with Trump’s global market may, for Southeast Asia, put notions of true sovereign AI out of the picture for now.
The slim prospect of a national AI stack, however, doesn’t annul Southeast Asia’s chances of asserting agency over this technology. ASEAN member states have a keen disposition towards regional multilateralism, which could emerge as a powerful convening force for tech development. Singapore has released its regionally tailored SEA-LION model, trained on a data corpus sourced from languages including Thai, Vietnamese, Tamil, and Indonesian. A proposed ASEAN Language Repository would elevate these efforts to a regional level through an open-access platform for member states to contribute their own linguistic and multimodal datasets. The new Digital Economy Framework Agreement between ASEAN states breaks ground as the first region-wide agreement on digital governance, harmonizing its status as an interoperable seedbed for digital growth.
By taking advantage of regional fora such as ASEAN, Southeast Asian countries can disperse the overwhelming costs of pursuing AI sovereignty and assert a unique leverage over their digital futures. Though perhaps less secure than developing a domestic stack, regional multilateralism could be the key to ASEAN states holding their own in the shifting terrain of AI geopolitics—and its rapidly thickening binary.
Dive Deeper with ASPI:
Read Wendy Cutler and Shay Wester’s recent op-ed for the Straits Times, “Cloud Computing is the Missing Link in Regional AI Ambitions.”
Watch Lizzi C. Lee moderate a virtual conversation on “China’s Open-Source AI Revolution,” part of a broader video series, China’s DeepSeek Moment, by ASPI’s Center for China Analysis.
Read Danny Russel’s report with Emily Ratté, “Defense or Diffusion? Open Source AI in U.S.-China Competition.”