Building AI systems that reason within the rules they operate under.
Applied research · AI for high-stakes decisions
Building AI systems that reason within the rules they operate under — across law, regulation, and professional practice.
Mission
Most applied AI optimises for prediction — what is likely to happen. The institutions making consequential decisions are rarely asking only what is likely. They are asking what is permitted, what is required, and what can be justified on the record. Those are questions of structured reasoning under uncertainty, and they do not reduce to pattern-matching.
The Initiative builds systems that make the steps of a difficult judgment explicit — the factors in play, the standards that bind them, and the points genuinely in dispute — so a human decision-maker can see the reasoning and remain accountable for it. The machine structures the problem; the person decides.
Aligned by design means the constraint comes first. Whether the governing frame is law, regulation, clinical guidance, or professional duty, accountability, transparency, and fairness belong in the reasoning structure of the system itself — not bolted on as a policy layer after deployment. A system that cannot show its working, or that quietly substitutes its own judgment for one a person is answerable for, is not fit for the institution that relies on it. Ours are built the other way round.
The rules that govern a decision are a constraint to reason within — not a variable to optimise around.
The Work
Approach
The Initiative is being built deliberately. Like the most durable public institutions, its foundation — doctrinal, methodological, and institutional — is laid before deployment, not after. That groundwork is the founder's doctoral research; a first release follows its completion.
The pace is the point. Systems that institutions will rely on for consequential decisions have to earn legitimacy before they scale, not retrofit it once they have. Building the social, legal, and political base in advance is what lets the work endure beyond a single moment or mandate — which is the only standard worth designing to.
Advisory
Governance frameworks
Structures for deploying AI in high-stakes settings — public administration, health, finance, and regulated industry — designed to hold against legal and professional scrutiny, with privacy and data-protection obligations (GDPR, PIPEDA, HIPAA) treated as design constraints rather than compliance afterthoughts.
Transparent decision architecture
Systems where the reasoning is explicit and auditable — ethics built into the structure of the decision, not added as a policy layer. Where data is sensitive, work begins from zero-data and synthetic prototypes before any live deployment.
Independent expert review
Assessment of systems and procurement — examining what a tool reasons over, and what it leaves to the person.
About
The Ethical AI Initiative is founded and led by Jeinis Patel — barrister, doctoral researcher in AI and Law at the University of London, and founder of the Initiative. The work began in law and extends to other rule-governed fields: building systems that institutions can stand behind.
Contact
For governance review, deployment guidance, or a conversation about a specific brief, write directly.