Research
Research, without theatre.
Privacy engineering, AI agents, scientific evaluation, and UK technology governance. We show enough to prove depth; we do not publish client systems, proprietary methods, or exploitable detail.
Research programme
The public edge of the work.
Our internal research is more detailed than the website. The public layer is deliberately selective: enough to show scientific understanding, not enough to hand over competitive intelligence, evaluation prompts, exploit paths, or client-specific architecture.
AI agents and evaluation
How to measure agentic systems before they are trusted with tools, memory, workflow autonomy, or customer-facing decisions.
Privacy-enhancing technologies
Differential privacy, selective disclosure, zero-knowledge proofs, secure computation, confidential computing, and where each primitive is actually useful.
Digital identity and age assurance
The technical difference between proving an attribute and building linkable identity infrastructure.
Scientific method for product claims
Borrowing from Turing, Popper, and modern AI-for-science: define the claim, test the failure mode, publish only what survives.
Science watch
Signals from top labs.
We track public work from AISI, OpenAI, Google DeepMind, the Alan Turing Institute, and adjacent research communities. We are especially interested in evaluated agents, AI-for-science loops, and the Turing line from computation to operational tests.
AISI
Frontier evaluation is becoming infrastructure
The UK AI Security Institute frames advanced-AI governance as a scientific measurement problem: capabilities, risks, mitigations, and repeatable tests.
OpenAI LifeSciBench
Scientific agents need research-grade benchmarks
Real scientific work involves evidence handling, uncertainty, experimental design, validation, and communication; not just clean quiz answers.
OpenAI + Molecule.one
Closed-loop AI science now reaches wet-lab feedback
Near-autonomous chemistry work shows the pattern we care about: propose, test, measure, replicate, and keep expert judgement in the loop.
Google DeepMind
Evolutionary search is becoming an engineering primitive
AlphaEvolve is a useful public reference point for measurable recursive improvement: generate candidates, score them, retain what works.
Alan Turing Institute
UK AI research still starts with rigorous definitions
The Turing lineage matters: computation, evidence, decidability, and operational tests before grand claims.
- 16 July 2026·8 min read·Privacy-Enhancing Technologies·Federated Learning
Federated Computing: Deploying AI on Data You Cannot Move
Federated learning, differential privacy and trusted execution environments let organisations train and run AI on sensitive data that regulation or competition keeps in place. What has shipped, what the regulators say, and where the claims go beyond the evidence.
Read article - 6 May 2026·6 min read·Digital ID·UK Policy
UK Digital ID: Ten Safeguards Parliament Must Legislate
Our response to the Cabinet Office consultation on digital identity (CP 1498): support in principle, with ten safeguards we want in primary legislation.
Read article - 8 April 2026·6 min read·Methodology·VPN
How We Score VPN Evidence: TheVPNMatrix Methodology Explained
How TheVPNMatrix rates the evidence behind its VPN assessments: the eight-component Evidence Quality Score and the 28 criteria it is applied to.
Read article - 19 March 2026·4 min read·GDPR·Privacy
What Is Privacy-by-Design and Why It Matters for UK Businesses
Privacy by design is a legal duty under Article 25 of the UK GDPR. This guide sets out the seven original principles and how a UK business can apply them.
Read article - 15 March 2026·5 min read·GDPR·Startups
A Practical Guide to GDPR Compliance for UK Startups
What a UK startup has to do under the UK GDPR, with links to the law. Reviewed in November 2026 for the Data (Use and Access) Act 2025.
Read article - 10 March 2026·5 min read·AI Governance·Regulation
AI Governance: What UK Businesses Need to Know in 2026
What the EU AI Act and the UK's regulator-led approach require of UK businesses, and how to set up governance for the AI systems you use.
Read article - 12 February 2026·6 min read·Facial Recognition·Civil Liberties
Facial Recognition in the UK: Built Before Parliament Authorised It
Police facial recognition is being bought and deployed before there is a specific law for it. A summary of our response to the Home Office consultation, with sources.
Read article
