Research that ships.
Sireto operates at the intersection of artificial intelligence, software engineering, knowledge representation, legal informatics, formal methods and distributed systems. We explore emerging ideas, build reference implementations and turn the approaches that work into production technology.
Six fields, one problem.
Making rules explicit enough for machines, and accountable enough for people, needs all of them at once.
The objective is not AI experimentation for its own sake.
It is to build systems that can be understood, tested and trusted.
- Artificial Intelligence
- Software Engineering
- Knowledge Representation
- Legal Informatics
- Formal Methods
- Distributed Systems
What we are working on.
Our research follows the problems NormLogic has to solve. Each theme has a reference implementation, an evaluation, or both.
Reliable requirement extraction
How far can AI-assisted extraction of obligations, permissions, conditions and exceptions be made measurable and repeatable? We build datasets and benchmarks to find out, and to know when a model is wrong.
Canonical regulatory models
Representations expressive enough for legal nuance yet precise enough to execute. We work with LegalRuleML, Akoma Ntoso and our own domain-specific languages.
Agents that operate against norms
Architectures where an agent’s actions are checked against explicit structured regulatory models rather than against retrieved text, and where each decision carries its justification.
Temporal and versioned regulation
Which version of a rule applied at a given moment, what changed in an amendment, and how that change propagates through dependent rules and systems.
Auditable reasoning
A chain of evidence from source provision to interpretation, norm, reasoning step and decision, so a conclusion can be inspected, challenged and reproduced.
Human–AI modelling workflows
How domain experts and AI systems build and maintain structured regulatory models together, which is the question NormFlow exists to answer.
From idea to production.
- 01
Explore
Study an emerging idea against a real regulatory problem, with a clear hypothesis about what would make it useful.
- 02
Build a reference implementation
Make the idea concrete enough to test, evaluate it systematically, and publish what we can.
- 03
Ship what works
Turn the approaches that hold up into production technology inside NormLogic or in client systems.
Working on a hard problem in this space?
We collaborate with researchers, regulators, EU research and funding partners and organisations on Regulatory Engineering, legal informatics and trustworthy agent architectures.