Nobody wants to admit that their billion-dollar strategic decisions are being guided by algorithms they don’t control, can’t audit and barely understand. Yet that’s exactly where Southeast Asia’s C-suite finds itself in late 2025. Over 70% of companies in Asia Pacific have adopted AI in at least one business function, with AI investment expected to surpass $110 billion by 2028. But whose AI, exactly?
When DBS Group CEO and Director Tan Su Shan announced in November 2025 that AI generated over S$750 million in economic value in 2024 – with projections exceeding S$1 billion for 2025 – it became the poster child for Southeast Asia’s AI revolution. But the models powering these gains? Overwhelmingly foreign-built, foreign-controlled and subject to geopolitical forces that Southeast Asian boardrooms can’t influence.
In other words: ASEAN’s most sophisticated institutions are building their competitive advantage on foundations that belong to someone else. And heading into 2026, that dependency could become a boardroom-level strategic vulnerability.
The Black Box Problem Nobody’s Solving
AI systems that operate as ‘black boxes’ create existential risks in regulated sectors like banking. When a credit algorithm denies a loan or flags a transaction as suspicious, can the bank explain why? Often, no. The model processed thousands of variables in milliseconds and reached a conclusion, but the reasoning remains opaque even to the institution deploying it.
When 97% Admit They Have No AI Security
Compromised data
Operational disruption
63% breached with NO AI governance policy
This isn’t just a compliance headache. It’s a liability crisis. Harvard Law School research shows just 11% of major corporations have explicit board or committee-level responsibility for AI oversight. If Western financial institutions with mature regulatory frameworks struggle with AI governance, what does that say about ASEAN banks with even less transparency?
When Indonesian banks use OpenAI’s models to automate credit decisions, they’re outsourcing governance to Silicon Valley. The model makes the call. The institution takes the liability.
Even Singapore’s financial institutions face this challenge. The Monetary Authority of Singapore’s December 2024 guidance on AI Model Risk Management acknowledges that banks must implement controls to prevent AI systems from generating unreliable outputs when confidence is low. Some may argue that this is more damage limitation than control.
The Geopolitical Vice Tightening into the New Year
The U.S.-China tech war has turned ASEAN into a contested battleground. Three events in May 2025 confirmed the AI rivalry entered a dangerous new phase: a Senate hearing on ‘Winning the AI Race,’ sweeping U.S. bans on Huawei’s AI chips and Trump’s Middle East chip diplomacy tour.
For ASEAN institutions, this creates an impossible choice. Washington’s AI Action Plan envisions exporting everything from chips to software standards, but only to nations signing onto America’s technology alliance. Meanwhile, China’s Premier Li Qiang emphasised creating a World AI Cooperation Organisation based in Shanghai.
Meaning: pick a side. Non-alignment is becoming untenable. Even when the strategic imperative is clear, execution remains elusive.
The Pragmatic Path Forward
Singapore’s approach emphasises consensus-building between government and industry, with voluntary frameworks like the Model AI Governance Framework and AI Verify toolkit rather than mandatory legislation. The principle: sovereignty isn’t about owning every layer of the stack. It’s about maintaining strategic autonomy where it matters most through flexible, principles-based governance.
DBS launched DBS-GPT for 5,000 employees, built guardrails and accepted that perfect independence isn’t the goal. Useful independence is. Meanwhile, Microsoft invested $1.7 billion in Indonesia to train over 840,000 people, creating talent that could reduce long-term dependency.
The uncomfortable truth: banks relying on external AI vendors inherit operational exposure to systems they don’t regulate, and structural dependence on a small oligopoly controlling both models and computing infrastructure.
Yet total rejection isn’t viable. Southeast Asia’s total gross domestic product (GDP) could rise to between 13% and 18% (a value nearing US$1 trillion) by 2030, thanks to accelerated AI adoption. The winners in the future will acknowledge dependency whilst systematically reducing exposure in high-risk areas.
What Happens Next
ASEAN’s Guide on AI Governance remains non-binding with no enforcement mechanisms and that’s inadequate. What’s needed now is coordinated action where national research centres pool resources to train models jointly, avoiding prohibitive costs of going it alone.
The question for the foreseeable future isn’t whether ASEAN institutions will use AI – 78% of companies globally already do. It’s whether ASEAN will build the governance, technical capacity and regional coordination needed to avoid becoming permanent digital colonies of either Silicon Valley or Shenzhen.
DBS’s Tan captured it bluntly: “The proliferation of generative AI has been transformative for us”, noting a ‘snowballing effect’ of benefits. That snowball is real, but so is the dependency it creates.
Right now? That question remains uncomfortably open. And the window for answering it is closing fast.
