How Singapore Out-Thinks Its Size

Abstract

This chapter sets out the institutional landscape behind Singapore’s research and artificial-intelligence agenda as it bears on a European research or applied-AI operation deciding whether to establish here. It describes the National AI Strategy and the governance bodies that direct it, the formally voluntary AI governance instruments that function as procurement benchmarks, the national research-funding architecture and the institutions that deploy it, the government’s role as an early and demanding customer, the regulated-sector markets where applied AI has genuine demand, and the defence-innovation institutions considered strictly as a market and funder for dual-use technology. The institutions are named, the funding is given with primary sources, and the instruments are dated. The chapter is candid about the central trade-off: Singapore’s model is deliberate, well-funded, and state-shaped, and it rewards firms whose work aligns with national priorities more readily than it rewards firms seeking a purely market-driven environment. A worked example follows a German applied-AI firm through the landscape, and the chapter closes with the recurring mistakes European firms make in reading it. The analysis is current as of the date of publication; this is a fast-moving policy area, and several of the instruments and figures cited were issued or revised in 2025 and 2026.

7.1 The Strategy Behind the Frameworks

Earlier chapters have touched the same machinery from different angles. The regulatory sandbox described in Chapter 5, the research-and-development incentives in Chapter 6, the data-centre stewardship in Chapter 3: each is a piece of one national strategy, and each makes more sense once you have seen the whole. This chapter draws that strategy out and shows you the institutions behind it, so that a research or applied-AI operation can work out where, if anywhere, it fits.

The argument of the chapter is straightforward, and it is worth stating plainly before the detail arrives. Singapore has decided to be more useful to a serious research operation than a country of its size has any right to be. It has done this not by spending recklessly but by organising. There is a named strategy, a council that owns it, a ministry that runs it, delivery agencies that implement it, a multi-year research budget with published allocations, and a set of governance instruments that a foreign firm can read, align to, and point to in a tender. None of this is improvised. The coherence is the product.

The coherence is also the trade-off. A landscape this organised rewards firms whose work fits the organising logic and is a less natural home for firms that want a purely market-driven, private-risk-capital environment. Section 7.8 makes that case honestly, because the honesty is what makes the rest of the chapter worth trusting. For now, hold both ideas at once: the institutional coherence is real and valuable, and it comes with a shape that a European firm should understand before it commits.

7.2 The National AI Strategy and Its Governance

Singapore published its first National AI Strategy in 2019, and its second, the National AI Strategy 2.0 (NAIS 2.0), in December 2023, under the vision “AI for the Public Good, for Singapore and the World.”¹ NAIS 2.0 reframed AI from an opportunity into a strategic necessity and set priorities for the following three to five years across what the strategy describes as three systems and ten enablers.² The point worth holding is not the taxonomy but the fact of it: there is a single, published, government-owned statement of intent that a firm can read.

That statement has moved quickly since, and a firm should check the current position rather than rely on a year-old summary. Two recent changes matter. In February 2026 the government established the National AI Council, chaired by the Prime Minister, to provide strategic direction and drive the national AI agenda.³ In May 2026 it released an Update to NAIS setting out ten refreshed priorities, building on NAIS 2.0 and supporting the Council’s elevated ambitions.⁴ The pace of these changes is itself a fact about the environment. A European firm planning a multi-year research operation should treat the strategy as a live document, read the most recent version, and expect it to have moved again by the time the operation is running.

The governance that delivers the strategy sits across three layers, and it helps to meet them as actors rather than as an organisation chart. The Ministry of Digital Development and Information is the responsible ministry, setting and refreshing priorities.⁵ The delivery agencies implement: the Infocomm Media Development Authority is the lead agency for the digital economy and for AI governance, and the Personal Data Protection Commission administers the data-protection regime that AI systems must operate within.⁶ Above them, the National AI Council now provides the strategic steer.³

Here is the earned acknowledgement, and it is load-bearing for a firm deciding where to base a regulated-sector AI operation. Singapore was among the first jurisdictions in the world to articulate AI governance principles, publishing the first edition of its Model AI Governance Framework in January 2019 and updating it the following year, and it has continued to update its instruments as the technology changed.⁷ For a European firm, the practical value is not that Singapore is “ahead” in some abstract race. It is that the firm operates against a published, reasonably stable governance reference rather than a vacuum. A firm that has lived through the uncertainty of building AI products against rules that did not yet exist will recognise the value of that immediately.

7.3 The AI Governance Instruments: A Reference, Not a Cage

The governance instruments are best understood as a stack that grew as the technology did. The Model AI Governance Framework for Traditional AI came first, published in January 2019 and updated in 2020, offering private-sector organisations implementable guidance on the ethical and governance issues of deploying AI.⁷ When generative AI arrived, the Infocomm Media Development Authority and the AI Verify Foundation (the Authority’s wholly owned, not-for-profit subsidiary) proposed a Model AI Governance Framework for Generative AI in January 2024 and finalised it on 30 May 2024, structured around nine dimensions, from accountability and data through to incident reporting.⁸ Most recently, the Authority launched the world’s first Model AI Governance Framework for Agentic AI on 22 January 2026, with a revised Version 1.5 published on 20 May 2026 that runs to a four-pillar structure covering risk assessment, human accountability, technical controls, and end-user responsibility.⁹

Alongside the frameworks sits the testing infrastructure. AI Verify is an AI governance testing framework and software toolkit, available since 2022 and stewarded by the AI Verify Foundation, that lets an organisation assess its AI system against eleven internationally recognised governance principles and generate a report on the result.¹⁰ A related open-source toolkit, Project Moonshot, addresses the evaluation and red-teaming of large language models.¹¹ On the data side, the Personal Data Protection Commission has issued advisory guidelines on the use of personal data in AI recommendation and decision systems, setting out how it intends to interpret and enforce the Personal Data Protection Act in an AI context.¹²

Now the point that earns its place, and the honesty that has to come with it. These instruments are formally voluntary. Compliance with the Agentic AI framework, for example, is voluntary, though organisations remain legally accountable for what their agents do under laws that do apply.⁹ The advisory guidelines are not themselves binding; they describe how the regulator will read the binding statute.¹² A firm that concludes from “voluntary” that the instruments can be ignored has misread the environment. In practice they function as benchmarks. A European firm that aligns its AI systems to the relevant framework, and can show the result of an AI Verify assessment, holds a recognised assurance position when it sells into a regulated sector or into government, the subjects of Sections 7.5 and 7.6. Alignment is not free; it asks for documentation, testing, and governance process that a small firm may not have built. But the cost buys a credential that the buyers in this market actually recognise, which is more than can be said for a self-asserted claim of responsibility. Chapter 4 set out the data-protection regime these instruments sit on top of; the reader should treat the two together.

7.4 The Research-Funding Architecture

Singapore funds research through a five-year national framework called Research, Innovation and Enterprise, now in its eighth iteration. The current plan, RIE2030, was released on 5 December 2025 with a budget of SGD 37 billion, a thirty-two per cent increase on RIE2025’s SGD 28 billion, and it runs from April 2026.¹³ The plan amounts to roughly one per cent of gross domestic product, which places Singapore in the company of small advanced economies such as Sweden and Denmark on research intensity, and the National Research Foundation coordinates the framework across the institutions that deploy it.¹³ One feature of RIE2030 is directly relevant to a European firm: a new Singapore–Horizon Europe Complementary Fund, which signals an intent to connect Singapore’s research funding to the European programme that many European firms already work within.¹⁴

The institutions that spend this money are worth meeting individually, because a firm seeking a research partner is choosing among them, not among abstractions.

The Agency for Science, Technology and Research, known as ASTAR, is the principal public-sector research agency, and it bridges academia and industry across a set of research institutes spanning the biomedical sciences, the physical sciences, and engineering, among them the Institute of High Performance Computing, the Institute for Infocomm Research, and the Institute of Microelectronics, which are the ones an applied-AI or deep-tech firm is most likely to encounter.¹⁵ ASTAR is the body a firm partners with when it wants institutional research capacity rather than a single academic collaborator.

AI Singapore is the national programme that runs applied-AI engineering work and develops local-context AI capability; it appears repeatedly in the institutional record as the partner that adapts models to the Singapore and regional operating context, including in the defence partnership discussed in Section 7.7.¹⁶ For a firm whose value is in applied AI rather than foundational research, AI Singapore is often the more natural first contact than A*STAR.

The Digital Trust Centre is the national centre for trust technologies (privacy-enhancing technologies, AI assurance, and the like), hosted at Nanyang Technological University and funded by a SGD 50 million grant from the Infocomm Media Development Authority and the National Research Foundation, beginning on 1 October 2022.¹⁷ The Centre also houses the Singapore AI Safety Institute, and it runs research grant calls that pair industry problem statements with research institutions.¹⁷ For a European firm working in privacy technology or AI assurance, a substantial and growing category, the Digital Trust Centre is the institution whose remit most directly overlaps its own.

The universities, principally the National University of Singapore and Nanyang Technological University, carry out the foundational research and the commercial translation that sit alongside the agencies.¹⁵

The earned acknowledgement here is the most important credit in the chapter, and it should be stated without hedging. These are well-funded, internationally connected research institutions that partner seriously with firms doing genuine research. The funding is real, the institutions are real, and the partnership routes are real. This is the reason a European deep-tech operation can find research partners and co-funding in Singapore rather than merely office space. Section 7.10 sets out the conditions on that credit, chiefly that a firm without genuine local research substance will not find the co-funding it imagines, but the credit itself stands.

7.5 The Government as a Demanding Customer

Singapore’s government is not only a regulator and a funder. It is a buyer, and an early and demanding one, and that role creates a commercial opportunity that a European firm should understand precisely because it is easy to misread.

The Government Technology Agency, GovTech, is the agency that builds and procures digital and AI capability for the public sector, and it functions as an anchor customer and early adopter for technologies the public service then adopts at scale.¹⁸ The government has also built tools to put AI in the hands of public officers directly and is driving broad-based adoption across agencies.⁴ Alongside the platform work sit funding and access initiatives aimed at enterprises rather than agencies, among them the Enterprise Compute Initiative, a SGD 150 million programme that helps companies reach advanced AI tools, cloud compute, and engineering support through cloud-service partners.¹⁹

The commercial logic for a European firm runs through reputation. A government that has tested a firm’s product against its own standards, and adopted it, becomes a reference customer whose endorsement carries weight with the next buyer, in Singapore and across the region. Meeting a demanding public-sector standard once is itself a saleable credential.

The honesty the section requires is about where the opportunity is real and where it is not. Public-sector procurement favours suppliers who can meet government security and deployment requirements, which for sensitive workloads can mean on-premise or otherwise tightly controlled hosting rather than a purely cloud-hosted service. A foreign firm with no local presence and a product that cannot be deployed inside a controlled environment will find much of this market closed to it, not through protectionism but through the security architecture the workloads demand. The opportunity is real for a firm willing to build to those requirements and oriented away from firms expecting to sell a standard cloud product into government unchanged.

7.6 Applied AI in the Regulated Sectors

The clearest applied-AI markets in Singapore are the ones where regulation and government demand combine to create a market that would not otherwise exist at this density. This connects directly to the use-case analysis of Chapter 2: the regulated-and-applied AI use cases fit Singapore not by accident but because the institutions described in this chapter create the demand.

Financial services is the leading example. The Monetary Authority of Singapore has driven industry initiatives that put applied AI to work on concrete regulatory problems, among them Project MindForge and the Veritas initiative on the responsible use of AI in the financial sector, and the COSMIC platform, operated with commercial banks, for sharing information to counter money laundering and related financial crime.²⁰ For an applied-AI firm, the significance is that the regulator has helped define the problems and the standards, which lowers the cost of finding a market and raises the value of meeting the standard. The Authority has also moved to set expectations directly: it has consulted on guidelines for managing the risks of AI use by financial institutions.²¹

Healthcare is a second example, organised around the public healthcare technology agency, Synapxe, which manages the technology backbone of the public health system and the deployment of analytics and AI within it.²² The public-service applications discussed in Section 7.5 are the third. In each case the pattern is the same: a regulated environment plus government demand produces a market with defined standards, which is precisely the environment in which a European firm’s investment in alignment and assurance pays back.

7.7 Defence-Adjacent and Dual-Use Innovation: The Institutional Market

This section treats Singapore’s defence-innovation institutions strictly as a market and a funder for dual-use technology. It describes who funds, who procures, and the precedent for European firms working with these institutions. It does not describe operational capability, and a European firm evaluating this market should hold the same discipline: the opportunity is commercial and institutional.

The institutions are the Ministry of Defence; the Defence Science and Technology Agency, which functions as the central procurement and systems-engineering body and deliberately acquires and adapts commercial and dual-use technology to control cost; and DSO National Laboratories, the national defence research-and-development organisation, which describes a workforce of more than 1,800 engineers and scientists.²³ Together they form a well-resourced institutional buyer and research partner.

The precedent that matters for a European reader is concrete and recent. On 20 March 2025, the Ministry of Defence, the Defence Science and Technology Agency, and DSO National Laboratories announced a partnership with the French firm Mistral AI to co-develop generative-AI models for decision-support applications, with support from AI Singapore in adapting the models to the local operating context.²⁴ Two features of that partnership are the instructive part for any European firm contemplating this market. First, a condition of the work was that the models could be deployed and managed on-premise within internet-separated environments, which the parties described as critical for defence.²⁴ Second, the partner chosen was a European firm offering open-weight models that could be adapted, rather than a closed cloud service. The shape of the opportunity follows from those two features: this is a market for firms whose technology can be deployed inside controlled, sovereign environments and adapted to local requirements, and it is effectively closed to firms that can only offer a hosted, externally controlled product.

The earned acknowledgement is that these are serious, well-resourced research institutions that have already partnered with a European firm on real work. That makes the market real and creditable rather than notional. The barriers in Section 7.10 are real too, among them security clearance, deployment architecture, and local presence, but the precedent stands.

7.8 The State-Shaped Character: An Honest Account

This is the section that keeps the chapter honest, and a reader who has come this far is owed it plainly.

Singapore’s research-and-innovation model is deliberate, well-funded, and state-shaped. The strategy is set centrally, the funding flows through national frameworks with published priorities, and the government is simultaneously regulator, funder, and customer. This produces the coherence the chapter has been describing, and the coherence is a genuine competitive advantage for the firm whose work aligns with it.

It also has consequences a firm should weigh before committing. Commercial viability in this environment correlates with relevance to public policy and national priorities more strongly than it does in a market where private risk capital sets the direction. A firm seeking the kind of purely market-driven, private-venture-funded environment found in parts of the United States, or in some respects in London, will find Singapore a less natural fit. The funding rewards alignment with national missions; the government-as-customer model rewards firms that build to public-sector requirements; the institutional partnerships reward firms whose research substance the institutions value. A firm that arrives expecting an open market for any well-made product, indifferent to national priorities, will be disappointed not because the system is closed but because it is pointed somewhere specific.

This is a choice competently executed, not a deficiency. A small country with no domestic market of consequence has decided to organise its research and innovation around national priorities and to make itself an unusually coherent and reliable place to do aligned work. For the firm whose work aligns, that is a strength. The honest advice is simply that a European firm should know which kind of environment it is entering, and should not assume it is entering the other kind.

7.9 A Worked Research-Operation Example

Consider a representative case, of the kind this book uses to make the abstract concrete. A German applied-AI company, call it a mid-sized firm with a strong product in document-intensive workflow automation for regulated industries, decides to establish a Singapore research operation to serve regulated-sector clients across Southeast Asia. It is not a foundational-research lab; its value is in applied engineering and in adapting its product to demanding regulatory environments. How does the landscape of this chapter actually bear on it?

On research partnership, the firm’s natural first contacts are AI Singapore, for applied-AI engineering and local-context adaptation, and the Digital Trust Centre, if its work touches privacy-enhancing technology or AI assurance.¹⁶¹⁷ A*STAR’s Institute for Infocomm Research is the route if the firm wants institutional research capacity for a harder technical problem.¹⁵ The firm should approach these as a partner with genuine substance to contribute, not as a grant applicant; Section 7.10 explains why that distinction decides the outcome.

On funding, the firm sits within the RIE2030 framework’s priorities rather than outside them, and the Singapore–Horizon Europe Complementary Fund is the feature most directly relevant to a firm already working within the European research programme.¹³¹⁴ The firm should not, however, build a business case that assumes co-funding; the funding rewards substance and alignment, and it follows a real research contribution rather than preceding it.

On the governance instruments, the firm’s product sells into regulated sectors, so alignment with the relevant Model AI Governance Framework and an AI Verify assessment are not optional refinements but the assurance position that makes the product saleable to the buyers in Sections 7.5 and 7.6.⁹¹⁰ The cost of building that position is part of the cost of entering the market, and it pays back precisely because the buyers recognise it.

On the government-as-customer opportunity, the realistic reading is that the public sector is a demanding reference customer the firm can aim at once it has a local presence and can meet deployment and security requirements, not a quick early win.¹⁸ If the firm’s workflow-automation product can be deployed in a controlled environment, the public sector and the regulated industries become reachable; if it cannot, much of this market is closed.

The honest summary for this firm is that Singapore is a strong fit. Its value is applied and regulated-sector-facing, which is exactly what the institutions reward; its European origin is an advantage rather than an obstacle, as the Mistral precedent shows; and its main risk is the temptation to over-index on grants rather than on building the substance and the assurance position that actually open the market.

7.10 The Eight Innovation-Landscape Mistakes European Firms Make

Treating the voluntary AI governance instruments as irrelevant. They are formally voluntary, but they function as procurement benchmarks. A firm that ignores them forgoes a recognised assurance position; a firm that aligns to them gains a credential the buyers in this market actually accept.

Assuming research co-funding without genuine local research substance. The funding rewards a real research contribution and alignment with national priorities. A firm that arrives expecting co-funding to subsidise an operation with no local research substance has misread the model and will not get it.

Misreading the government-as-customer opportunity as easy. The public sector is a demanding, security-conscious buyer, not a soft early market. The opportunity is real for a firm that builds to its deployment and security requirements and remote for a firm that expects to sell a standard cloud product unchanged.

Treating the state-shaped model as a free market. Commercial viability here correlates with relevance to national priorities. A firm that ignores the national missions and expects an open market indifferent to them will find the environment less responsive than it expected.

Over-indexing on grants rather than capability. Grants follow substance; they do not substitute for it. A firm that organises itself around chasing funding rather than building a genuine capability and assurance position has the causation backwards.

Misjudging the dual-use market’s procurement reality. The defence-innovation market is real and has taken European partners, but it requires deployment inside controlled, sovereign environments, adaptability, and local presence. A firm offering only a hosted, externally controlled product is effectively excluded regardless of product quality.

Failing to align with national research priorities when seeking partnership. The research institutions partner most readily with firms whose work serves the priorities the institutions are funded to advance. A firm seeking partnership should know those priorities and speak to them, not present an undifferentiated capability.

Assuming the AI strategy is static. The strategy and its governance have moved fast and recently: a new council in February 2026, a refreshed strategy update in May 2026, and a new agentic-AI framework in January 2026 revised in May 2026.³⁴⁹ A firm that plans against a year-old version of the landscape is planning against a position that has already changed.

7.11 Conclusion

Singapore’s research-and-innovation institutions are the clearest expression of the strategy that runs through this whole book: a small country deciding to be more useful, more reliable, and more institutionally coherent than its size would predict. The named strategy, the council that owns it, the ministry and agencies that run it, the multi-year research budget with published allocations, the governance instruments a firm can align to and point to, the government as an early and demanding customer, and the defence-innovation institutions that have already taken a European partner: these are not a collection of separate programmes but a single, deliberate design.

The design has a shape, and the chapter has tried to show it honestly. It rewards firms whose work aligns with national priorities and is a less natural home for firms that want a purely market-driven environment. For the European research or applied-AI operation whose work does align, and the regulated-sector, applied, assurance-conscious firm is exactly that case, this coherence is one of the strongest reasons to be present, and the European origin of the firm is an advantage rather than an obstacle.

The remaining chapters turn from the strategy to the practicalities: the property, the talent, the entity, and the multi-year reality of operating here. The institutions described in this chapter are the reason those practicalities are worth working through.

References

Declarations

Competing interests: The author is a licensed real estate agent (Council for Estate Agencies, Singapore) affiliated with OrangeTee & Tie Pte Ltd, and a Singapore Mediation Centre-accredited mediator. The author has commercial interests in industrial and commercial real estate transactions facilitated through OrangeTee & Tie. These interests are openly disclosed. The analysis in this chapter has been written to be useful to the reader irrespective of whether the reader subsequently engages the author’s transactional services.

Funding: This work received no external funding.

Methodology: This chapter is built on Singapore government primary sources: the Smart Nation Singapore and Ministry of Digital Development and Information statements of the National AI Strategy and its governance, the Infocomm Media Development Authority and AI Verify Foundation publications of the AI governance frameworks and testing toolkits, the National Research Foundation and Economic Development Board announcements of the Research, Innovation and Enterprise 2030 plan, the Digital Trust Centre and host-university materials, the Government Technology Agency and Monetary Authority of Singapore programme materials, and the Ministry of Defence, Defence Science and Technology Agency, and DSO National Laboratories public announcement of the named European defence-AI partnership. Each substantive factual claim was verified against the relevant primary source at the time of drafting. Several governance instruments and budget figures cited were issued or revised in 2025 and 2026; where a specific figure could not be verified against a primary source it has been omitted rather than reproduced.

Currency of analysis: The analysis is current as of the date of publication. Singapore’s AI strategy, governance instruments, and research-funding allocations are revised frequently and on announced trajectories; the National AI Council, the May 2026 strategy update, the agentic-AI framework, and the RIE2030 budget cited here all date from 2025 or 2026. A reader acting on this chapter should confirm the current position of each instrument and figure before relying on it.

About the Author

David Hoicka is a Singapore-licensed real estate agent (Council for Estate Agencies) affiliated with OrangeTee & Tie Pte Ltd, with a specialisation in industrial and commercial property for European inbound investment. He is also a Singapore Mediation Centre-accredited mediator, a civil engineer (Bachelor of Science, Massachusetts Institute of Technology), and the founder and publisher of Singapore Mediation Solutions, an academic publisher registered with Crossref (DOI prefix 10.66404) and with the National Library Board of Singapore. He has lived in Singapore as a permanent resident for over twenty-one years.

Scholarly identifiers: ORCiD 0000-0001-9082-0720; Wikidata Q137455251; ISNI 0000 0005 2886 676X; Google Scholar profile available.

About the Publisher

Singapore Mediation Solutions is an open-access scholarly publisher specialising in practical and analytical works for cross-border commercial practitioners with a focus on Asia-Europe industrial and commercial relations. Singapore Mediation Solutions is registered with Crossref (DOI prefix 10.66404), is a Singapore publisher with NLB-assigned ISBNs, and deposits all works in Zenodo for permanent open-access availability and in OCLC WorldCat for library catalogue accessibility.

Confidential Consultation

Readers who would like to discuss establishing a research or applied-AI operation in Singapore in confidence may contact the author directly. The preferred channels are Signal and Telegram for confidentiality and ease of cross-border communication. Direct email is also available. Contact details are listed on datascienceai.org. Initial consultations are conducted without obligation; the author’s role as principal advisor and the relationship to OrangeTee & Tie transactional execution are set out in a written engagement letter before any onward referrals are made.

---

Chapter DOI: 10.66404/de.b5.ch7 (to be assigned upon Crossref deposit) Zenodo deposit: pending Published by Singapore Mediation Solutions, Singapore Open access under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0)