Flint advises international businesses and investors on policy, politics, regulation and competition.
Flint advises international businesses and investors on policy, politics, regulation and competition.
Quick links
Sectors
Follow Us
AI-Pac is Flint’s digest providing our take on major AI policy and regulatory developments in Asia-Pacific. In a fast-moving environment, this newsletter is designed to help you stay on top of the regulatory agenda and consider what it means for your business. Please see here for our previous blogs, spanning 2024 to 2026. In each blog, we set out:
In May and June, there was continued political focus on AI sovereignty and resilience across Asia-Pacific. US restrictions on access to Anthropic's latest frontier models intensified government concerns around strategic dependence on foreign-owned AI capabilities. Meanwhile, the regulatory environment for AI continues to evolve rapidly around the region.
In this digest:
Questions around access to advanced foundation models escalated in June, when the US government imposed export controls on Anthropic’s Claude Fable 5 and Mythos 5 models, citing national security concerns. The controls were lifted on 30 June after Anthropic strengthened its safeguards and committed to deeper collaboration with the government, although access to Mythos 5 remains limited to Project Glasswing partners. This was not a one-off incident — OpenAI reportedly staggered the release of its latest models at the US government's request.
This episode has given fresh impetus to AI sovereignty concerns across the region. In several APAC countries, reliance on foreign AI providers is now seen by some stakeholders as a national resilience and economic security risk. The debate on AI sovereignty is particularly advanced in Korea and India. In Korea, the Deputy Prime Minister stressed that strategic autonomy requires the ability to develop and operate AI independently, and warned of the risks of long-term reliance on foreign models for critical national infrastructure and industrial security. The Indian government is reportedly set to take a small stake in Sarvam, a homegrown lab building indigenous foundation models, in what would be its first direct shareholding in a domestic AI developer.
Notably, the debate is no longer confined to governments, with industry and civil society also weighing in. In India, prominent investors and executives have called for greater investment in the domestic AI stack and greater use of open-source alternatives, while in Korea, the CEO of Upstage described AI as a strategic national asset and stressed the need to become self-reliant.
For non-domestic businesses, as we noted in our earlier analysis of sovereign AI in APAC, the risk is that these narratives translate into concrete policy levers, such as data localisation obligations, preferential policies favouring domestic champions, or restrictions on foreign AI models and cloud providers in critical sectors. With the meaning of AI sovereignty still contested, businesses also have an opportunity to position themselves as trusted partners in advancing national AI ambitions, whether through local infrastructure investment, skilling initiatives that strengthen the local talent pool, or contributing technical expertise to sovereignty debates that remain unsettled.
Southeast Asian governments continued advancing their AI governance agendas. In Malaysia, the government is accelerating work on an AI Bill ahead of Cabinet consideration in July, while preparing for the National AI Office to be officially institutionalised to oversee and coordinate AI affairs in the country. Extensive industry engagement is continuing ahead of the Bill's planned tabling in Parliament by year-end. In Indonesia, two Presidential Regulations on AI remain in the approval process following further industry consultations, with finalisation targeted for 2026. Meanwhile, Thailand advanced work on a new draft AI Act, proposing a risk-based approach informed by the EU AI Act.
A common theme across all three markets is that governments are looking to established international frameworks — particularly the EU’s risk-based approach — as a practical starting point for AI regulation, rather than designing entirely new regimes from first principles. At the same time, officials in all three countries privately indicate there is little appetite for transplanting overseas regimes wholesale. Instead, the consistent message — both publicly and in private — is that international approaches must be adapted to local conditions.
For businesses, this is an important window to help shape emerging AI governance frameworks before they are finalised. Southeast Asian governments continue to actively seek practical input from industry as they work through implementation questions, for instance Thailand's ongoing consideration of copyright protections and the use of copyrighted material for AI training. Businesses should focus their engagement on what a more locally appropriate, context-specific approach would look like in practice.
Singapore has moved quickly to establish itself as one of the first jurisdictions developing governance approaches specifically for agentic AI. Over May and June, the Infocomm Media Development Authority updated its Model AI Governance Framework for Agentic AI with new best practices and case studies; published a discussion paper examining legal responsibility and accountability for AI agents; and, together with GovTech, the Cyber Security Agency of Singapore (CSA) and Google, released a (Flint-authored) white paper on its AI Agents Sandbox — a global-first initiative that tests agentic AI systems in real-world settings to generate practical insights that can inform how they are developed, deployed and governed.
The pace of work underscores Singapore’s ambition to shape emerging global norms for agentic AI governance. Notably, the Model AI Governance Framework for Agentic AI was updated just four months after its initial release in January 2026, reflecting Singapore's iterative approach to governance. Equally notable was Singapore's decision to publish a discussion paper on agent liability despite many of the underlying legal questions remaining unresolved. Rather than seeking to provide definitive answers, the paper aims to equip policymakers with an early understanding of the novel legal challenges posed by increasingly autonomous AI systems.
More broadly, these initiatives reinforce Singapore’s view that effective AI governance is best developed through practical experimentation and close public-private collaboration. Looking ahead, as other jurisdictions begin considering governance approaches for agentic AI, it remains to be seen whether Singapore's practical and implementation-oriented approach will gain wider traction across the region. There is an opportunity for businesses to draw on Singapore's AI Agents Sandbox as a reference point, particularly in how they can similarly contribute practical expertise and collaborate with governments early in the policymaking process.
Sector-specific regulatory frameworks for AI proliferated across the region in May and June, most notably in financial services, legal services, healthcare, and public services. Financial services regulators were the most active: China’s National Financial Regulatory Administration released its first dedicated guidelines on AI in the banking and insurance sector; Korea’s Financial Services Commission introduced revised AI guidelines for the financial sector; India’s central bank released draft guidance on model risk management. Beyond finance, Hong Kong’s judiciary plans to consult the legal profession on draft guidelines for AI use, Indonesia signalled plans to draft regulations on AI in healthcare, and Australia established an AI Review Committee to scrutinise high-risk public sector deployments.
An emerging trend is that governments are turning to sectoral rules as the fastest available route to governing AI. In part, this reflects a recognition that sectoral regulators are often best-placed to respond to AI risks in their sectors, with both the context-specific understanding and the existing powers necessary to take action. For example, a central bank can restrict how regulated entities use AI in credit decisions by issuing a circular, without needing to introduce new legislation. By contrast, horizontal frameworks tend to take much longer, requiring legislative passage and consensus-building across multiple stakeholders.
Businesses will need to grapple with an increasingly fragmented compliance landscape across markets and sectors. As sectoral instruments move from draft guidelines to implementation far more quickly than horizontal legislation, businesses will have less time to adjust once new requirements are introduced. Businesses should map which sectoral regimes apply to their AI deployments across markets, and begin preparing positions on the specific questions that sectoral regulators are now addressing.
This blog was written by Ewan Lusty, Partner, Ethan Teo, Manager, and Sarah Lim, Manager. Ewan, Ethan, and Sarah are based in Singapore. Over six years of working for Flint in Europe and Asia, Ewan has advised leading developers and deployers of AI on how to engage impactfully with governments on AI policy and regulation. Ethan previously worked at Singapore’s Smart Nation Group under the Prime Minister’s Office and has interned with Grab, Twitter, and Carousell. Over the past two years at Flint, he has advised multiple clients on AI governance and regulation, including supporting a government–industry partnership sandbox initiative on agentic AI. In Sarah’s previous role at Singapore’s Ministry of Digital Development and Information (MDDI), she coordinated whole-of-government policy positions on AI governance under Singapore’s National AI Strategy 2.0.