What evidence should institutions require before increasingly capable AI systems are deployed in consequential settings?
Dr Shyam Krishna
Research
Research
AI safety, governance and strategic technology
My research examines how increasingly capable AI systems reshape institutions, infrastructure and the distribution of power. I focus particularly on the evidence, governance arrangements and preparedness institutions need as AI systems become more autonomous, interconnected and embedded in consequential settings.
My work spans AI safety and evaluation, assurance, critical infrastructure, agentic systems, human agency, strategic dependencies and international AI governance. Across these areas, I combine policy research with foresight, tabletop exercises, ecosystem mapping, sociotechnical analysis and mixed-methods research.
Research agenda
Where my work sits
Five overlapping areas structure my current research agenda.
01
AI Safety, Evaluation & Assurance
Understanding capability and building evidence for deployment
Research on emerging AI capabilities, failure modes, evaluation and the standards, assurance mechanisms and institutional evidence needed to govern increasingly capable systems.
02
AI, Critical Infrastructure & Agentic Systems
Governance when AI becomes operational
How autonomous and multi-agent AI systems affect resilience, accountability, escalation and operational decision-making across critical infrastructure and other consequential settings.
03
AI & Human Agency
Preserving meaningful human agency
How AI systems reshape human judgement, expertise, autonomy and the capacity to act, particularly as decision-making becomes increasingly mediated or delegated to automated systems.
04
Geopolitics, Sovereignty & Global Governance
Who controls capability, infrastructure and governance
How compute, semiconductors, cloud infrastructure, models and supply chains shape strategic dependency, while governments and international institutions build capacity to govern AI across different contexts.
05
Foresight & Preparedness
Testing institutions against uncertain futures
Using scenarios, tabletop exercises, ecosystem mapping and futures methods to examine where governance arrangements may fail and how institutions can prepare for uncertain technological change.
Research foundations
Methods
How I research emerging technology
Different research problems require different forms of evidence. My work combines conventional policy research with methods designed to examine technological systems while they are still evolving.
Institutions
Policy & institutional analysis
Examining governance arrangements, regulatory capacity, institutional responsibilities, incentives and implementation constraints.
Anticipation
Foresight & scenarios
Exploring plausible technological and institutional developments without assuming that a single technological future is predetermined.
Preparedness
Tabletop exercises
Stress-testing coordination, escalation, accountability and response arrangements through simulated AI-related events and crises.
Systems
Ecosystem mapping
Mapping relationships between firms, infrastructure, institutions, capabilities, supply chains and policy environments.
Evidence
Mixed-methods research
Combining interviews, qualitative analysis, documentary research, quantitative evidence and stakeholder engagement.
Sociotechnical
Sociotechnical evaluation
Studying AI systems alongside the institutions, incentives, social relationships and distributions of power within which they operate.
Across the agenda
Questions running through my research
These questions connect work that spans different technologies, sectors and institutional contexts.
How does greater AI autonomy change where responsibility, oversight and accountability should sit?
Where do dependencies on models, infrastructure and suppliers create new systemic or strategic risks?
How can institutions prepare for capabilities and failure modes that remain uncertain or difficult to forecast?
How can AI governance preserve meaningful human agency as systems become more capable and deeply embedded?
How can countries retain meaningful governance capacity when technological capability and infrastructure are concentrated elsewhere?
Intellectual foundations
From data justice to AI resilience
My earlier research focused on digital identity, platform labour, datafication, surveillance, technology for development and data justice. My wider background also spans ethnography, telecommunications, network security and Internet of Things technologies.
Across both my earlier and current research, the central concern is how technological systems redistribute agency, risk, visibility and responsibility — and how institutions can respond when those distributions become unequal or difficult to govern.