Fully Funded Singapore AI Safety Fellowship 2026

Artificial intelligence is no longer a technology of the future. It is the defining technology of the present — reshaping healthcare, education, finance, defence, creative industries, and the fundamental ways in which human beings communicate, decide, and understand the world around them. And with that transformation comes a question that the most serious thinkers in technology, policy, and philosophy are wrestling with urgently: how do we ensure that AI systems are safe, trustworthy, and aligned with the values and interests of humanity?

That question is not rhetorical. It is a genuine scientific and governance challenge — one that requires the best researchers in the world working on it, right now, with the seriousness and resources the stakes demand.

The Singapore AI Safety Fellowship 2026 is one of the world's most exciting new opportunities for researchers ready to engage with exactly that challenge. This fully funded three-month fellowship brings outstanding technical researchers and AI governance specialists from every country in the world to Singapore — one of Asia's most advanced and internationally connected cities — to conduct cutting-edge AI safety research under the mentorship of internationally recognized experts, with computing resources that most university laboratories can only dream of.

The fellowship runs from 21 September to 4 December 2026. There is no application fee. The deadline is 10 July 2026 — just days away. If you are a researcher with a strong technical background and a genuine passion for ensuring that AI development goes well for humanity, this fellowship was designed for you.

The AI Safety Challenge: Why This Research Matters Now

To fully appreciate the significance of the Singapore AI Safety Fellowship, it helps to understand the intellectual and practical landscape it operates within — and why the work of AI safety research has become one of the most urgent scientific priorities of our era.

The rapid advancement of AI capabilities over the past several years has been genuinely extraordinary. Large language models, multimodal AI systems, reinforcement learning agents, and a growing ecosystem of AI-powered applications have demonstrated capabilities that would have seemed implausible even five years ago. These systems can write, reason, code, create, and in certain narrow domains outperform the best human experts. And the trajectory of development — with each new generation of AI systems surpassing its predecessors — shows no signs of slowing.

This progress is genuinely exciting. The potential benefits of advanced AI — in accelerating scientific discovery, improving healthcare outcomes, making education more accessible, and solving complex engineering problems — are enormous and real. But the same capabilities that make AI systems powerful also create risks that deserve serious attention. An AI system that is capable but not reliably aligned with human values, that pursues objectives in ways that produce harmful side effects, or that is deliberately misused by malicious actors poses risks that scale with the system's capability.

AI safety research is the scientific discipline dedicated to understanding and addressing these risks — to developing the technical methods, the evaluation frameworks, the governance structures, and the policy instruments that can ensure AI systems are not just powerful but genuinely safe, trustworthy, and beneficial. It is a field that sits at the intersection of machine learning, mathematics, philosophy, cognitive science, and public policy — and it is a field in urgent need of more talent, more resources, and more international collaboration.

The Singapore AI Safety Fellowship is a direct investment in all three of those needs.

About the Singapore AI Safety Institute and the Fellowship

The fellowship is hosted by Singapore's AI safety research community — part of a growing national commitment to positioning Singapore as a global hub for responsible AI development and governance. Singapore has been among the most proactive countries in the world in developing institutional frameworks for AI safety and governance, recognizing that the risks and opportunities of advanced AI are global in nature and require international collaboration to address effectively.

The fellowship program reflects Singapore's distinctive approach to AI: technically rigorous, internationally engaged, and oriented toward both the scientific frontier and the practical policy implications of AI development. By bringing together researchers from around the world — creating what the program explicitly describes as a bridge between Eastern and Western AI safety research communities — the fellowship embodies the conviction that AI safety is too important to be addressed by any single national or cultural tradition alone.

Fellows join a research community that combines deep technical expertise in machine learning and AI systems with sophisticated thinking about AI governance, policy, and the broader social implications of advanced AI. This interdisciplinary environment is one of the fellowship's most valuable features — it ensures that purely technical AI safety research is always conducted with awareness of its policy context, and that AI governance work is always grounded in genuine technical understanding.

The fellowship is residential and full-time. Selected fellows relocate to Singapore for the three-month period, working from a dedicated office workspace in central Singapore and embedded in a community of researchers, mentors, and collaborators who are among the most serious and accomplished people working on AI safety in the world today.

Singapore: The Ideal Environment for AI Safety Research

The choice of Singapore as the home for this fellowship is not arbitrary. Singapore occupies a unique position in the global AI landscape — and understanding that position helps explain why spending three months here, working on AI safety research, is a particularly valuable experience.

Singapore is one of the world's most technologically advanced and digitally sophisticated societies. Its government has invested heavily in AI research and development, in digital infrastructure, and in the regulatory frameworks that govern AI deployment — creating an environment where AI research is both practically relevant and institutionally supported. The country's strong tradition of evidence-based policymaking means that AI governance work done in Singapore is taken seriously by policymakers and has real potential to influence both national and regional AI policy.

Singapore also occupies a strategic position as the principal bridge between East Asian and Western AI research communities. As AI development becomes increasingly global — with major research centers in the United States, the United Kingdom, Europe, China, Japan, South Korea, and India all contributing to the frontier — the ability to work across these different communities, understand their different approaches and priorities, and build collaborative relationships that transcend geopolitical divides is an increasingly important professional capability. Singapore's cultural and institutional position makes it an almost uniquely effective place to develop that capability.

The city itself is an extraordinary place to live and work. Among the cleanest, safest, and most efficiently run cities in the world, Singapore combines world-class infrastructure with a fascinating multicultural character — a city where Chinese, Malay, Indian, and global Western cultures intersect in ways that are simultaneously vibrant and harmonious. Its food scene is legendary — consistently rated among the best in the world — and its compact geography makes exploring the city and its surroundings effortlessly convenient. For researchers arriving from other parts of Asia or from further afield, Singapore offers an immediately comfortable and stimulating environment.

September through December — the fellowship period — captures Singapore in a pleasant phase of its tropical year. While the equatorial climate means warmth year-round, the northeast monsoon season brings some refreshing rain and cooler evenings that make the city particularly pleasant for the focused intellectual work that research demands.

The Research Agenda: Ten Disciplines at the AI Safety Frontier

The Singapore AI Safety Fellowship is organized around ten interconnected research areas — each one a critical dimension of the broader challenge of making AI systems safe, trustworthy, and beneficial. Fellows work within one or more of these areas, developing research projects that push the frontier of understanding in their chosen domain.

AI Alignment is the foundational challenge of ensuring that AI systems pursue the objectives their designers intend — that they are, in a deep sense, trying to do what we want them to do. Alignment research explores the mathematical and computational frameworks for specifying objectives, the failure modes through which aligned systems can become misaligned, and the technical methods for maintaining alignment as AI systems become more capable.

Frontier AI Safety addresses the specific safety challenges posed by the most capable AI systems — the large-scale models at the cutting edge of capability development. As these systems demonstrate increasingly general capabilities, the safety challenges they pose become more complex and more consequential. Frontier AI safety research works to understand and address these challenges before they manifest in deployed systems.

AI Governance moves from the technical to the institutional — exploring the frameworks, standards, regulations, and international agreements that can effectively govern the development and deployment of AI systems. This is work at the intersection of computer science, political science, law, and international relations, and it is work that has become increasingly urgent as governments around the world grapple with how to regulate a technology they often do not fully understand.

Machine Learning Safety investigates the specific ways in which machine learning systems — the dominant paradigm of current AI development — can fail in safety-relevant ways. This includes work on adversarial robustness, out-of-distribution generalization, distributional shift, and the ways in which machine learning systems trained on historical data can produce harmful outcomes in deployment.

AI Security addresses the deliberate misuse of AI systems — by adversaries who want to exploit vulnerabilities in AI-powered systems, use AI to automate cyberattacks, generate synthetic disinformation, or weaponize AI capabilities in other harmful ways. AI security research develops the defenses and resilience mechanisms needed to protect against these threats.

Trustworthy Artificial Intelligence is a broader research agenda encompassing the full range of properties that make AI systems worthy of trust — including fairness, accountability, transparency, reliability, and privacy. This research area sits at the intersection of technical AI research and the social and ethical frameworks within which AI systems are evaluated and governed.

AI Policy translates technical AI safety insights into the language and frameworks of public policy — developing the evidence base, the analytical tools, and the policy recommendations that can inform effective government action on AI safety. This is research that requires fluency in both the technical realities of AI systems and the political and institutional dynamics of policy-making.

Risk Evaluation develops the methodologies and frameworks for assessing the risks posed by specific AI systems or classes of AI systems — enabling more rigorous, evidence-based decision-making about when and how to deploy AI technologies. Risk evaluation research draws on lessons from other high-stakes technology domains, including nuclear safety, aviation, and pharmaceutical regulation.

Interpretability investigates the mechanisms by which AI systems — particularly neural networks — arrive at their outputs, with the goal of making these mechanisms understandable to human researchers and developers. Interpretability research is foundational to many other aspects of AI safety: if we cannot understand why an AI system behaves the way it does, we cannot reliably predict or control its behavior in novel situations.

AI Verification develops the formal methods and empirical techniques for verifying that AI systems satisfy specific safety properties — for establishing, with mathematical or empirical confidence, that a given system will behave safely in specified conditions. This is some of the most technically demanding work in AI safety, drawing on advances in formal methods, program verification, and statistical theory.

The Computing Resources: USD 30,000 — A Research Accelerator

One of the most practically significant features of the Singapore AI Safety Fellowship — and one that genuinely distinguishes it from most academic research fellowships — is the provision of up to USD 30,000 in computing resources for eligible research projects. For AI safety researchers, this is an extraordinarily valuable resource.

Modern AI research, particularly research on frontier AI systems, is computationally intensive in ways that can make it prohibitively expensive for researchers without access to well-funded laboratories or industry partnerships. Training models, running large-scale experiments, and evaluating AI system behavior across diverse conditions all require substantial computational power — typically provided through cloud computing platforms or specialized AI hardware.

By providing fellows with up to USD 30,000 in computing credits, the fellowship removes one of the most significant practical barriers to high-impact AI safety research. It enables fellows to run the experiments their research plans demand, at the scale those experiments require, without the constant budget anxiety that constrains so much independent and academic research. This computing resource is not a token gesture — at current cloud computing prices, it represents access to genuinely serious experimental infrastructure.

Weekly Mentorship: Learning from the Best

The fellowship's commitment to weekly mentorship from experienced AI safety researchers is another feature that sets it apart from conventional research experiences. Research can be isolating — particularly when you are working on problems at the frontier of a field where the answers are genuinely unknown and the methodologies are still being developed. Regular, structured access to experienced researchers who have navigated similar challenges is one of the most effective ways to accelerate learning and improve the quality of research output.

The mentors connected to the Singapore AI Safety Fellowship are drawn from the international AI safety community — researchers with track records of published work, policy engagement, and practical experience with the specific technical and governance challenges that fellows are working on. These mentorship relationships are not merely advisory; they are substantive, ongoing engagements that shape the direction and quality of the research fellows produce during their three months in Singapore.

How to Apply

1. Visit the official Singapore AI Safety Fellowship page at aisafety.sg/programs/singapore-ai-safety-fellowship
2. Review the complete fellowship description, research areas, and eligibility requirements carefully
3. Prepare your research background information, proposed research direction, and supporting materials
4. Complete and submit the online application form through the official portal
5. Submit your application before the 10 July 2026 deadline — just days away