The hardest problems in AI are no longer in the lab
An independent institute bringing together governments, global enterprises and the people who have deployed AI where failure is not an option, to make it work inside the systems it now runs.
The models became general.
The deployments did not.
Closing that gap is our mission.
Ambition is not the shortage, and neither is capital. The shortage is deployment: making advanced AI work, and keep working, inside the real-world systems it was never designed for. That is the discipline the Institute exists to advance, led by the handful who have done it at scale.
Nine fronts. One discipline.
These are the domains that will decide how AI reshapes government and industry. On each, the view from those who have actually deployed it: what works, what breaks, and what to do next.
Complex AI
A model is a component; a deployed AI system is a complex adaptive system, complete with feedback loops, emergence and cascade risk. Complex AI is the discipline of governing it: traceable, auditable, accountable across competing objectives.
Enter the frontier
Sovereign AI
National models, sovereign compute and data residency: capability, not theatre.
Discover →
Agentic AI
AI that acts. Accountability, permissions and audit for autonomous systems.
Discover →
Quantum AI
The quantum frontier, and the post-quantum migration that cannot wait.
Discover →
Edge AI & Compute
Intelligence where the data lives, and the compute stack beneath it.
Discover →
Defence & Cyber AI
Assurance for the highest-stakes autonomy, from the battlefield to the fifth domain.
Discover →
Physical AI & Robotics
Embodied intelligence: the hardest deployment test of all.
Discover →
Health & Longevity AI
The deployment gap measured in lives: clinical assurance, adaptive models, the longevity economy.
Discover →
AI Policy & Regulation
The global rulebook as it actually stands, and how deployers navigate it.
Discover →From the Institute
The GCC AI Regulatory Tracker
Six Gulf states, four regimes, one interactive map: data protection, cross-border transfer, AI governance and free zones, every entry sourced to primary law. The finding that matters for anyone deploying in the region: every state now has enforceable data protection, and not one has a binding general AI law.
Open the tracker
The Gulf's AI Patchwork
Six states, nine data-protection rulebooks, six transfer regimes and not one binding AI law. The Gulf’s fragmentation, counted rather than asserted.
Read →
The Deployment Gap: Why AI Dies in Production
Most AI deployments fail, and the model is rarely the reason. The evidence from government and enterprise, why systems die in production, and a checklist for ministers and boards.
Read →
Who Decides What Counts as Trustworthy?
The UK wants to rank “trustworthy” news above the algorithmic noise. The goal is right and the instrument is wrong: certify the claim, not the masthead.
Read →Proof, not commentary.
Fellowship of the Complex AI Institute is by invitation: builders, governors and operators of real AI systems, appointed on the record of what they have deployed. The first five fellows are named, and the rest of the founding cohort follows this autumn.
Dr Rabimba KaranjaiAgentic AIACM SIGSOFT Distinguished Paper Award, 2025
Steve SuarezQuantum AIFormer Global Head of Innovation at HSBC and lecturer at MIT
Michael StattelmanDefence & Cyber AIArchitect of a vision model downloaded 1.3 billion times
Dr Young-Dall LeeAI Policy & RegulationThree presidential commissions, Republic of Korea
Sam BeniComplex AILed Tech Nation's UK Government Applied AI Programme
Non-profit. Independent. On the record.
The Institute is being established as a non-profit: all surpluses are reinvested in its mission. Published research is firewalled from advisory engagements.
One email a month on AI in the real world.
What shipped, what failed, and why. Unsubscribe any time.