Published research
Methods we publish.
Including the limits.
Argue with them.
Research notes from the Enterprise AI Lab. Each one sets out a working method and is explicit about what it cannot do — which is the part most methodologies leave out.
Research notes
What we have published.
These are working methods rather than thought leadership. They are written to be used by someone else’s team, and to be criticised by someone else’s regulator.
The unassigned workstream
Councils are being abolished to a statutory deadline and nobody owns the AI workflows, evidence or audit records. Five tests and a handover schedule.
Read the note Research noteRed-teaming high-risk AI under Article 9
A five-stage working method, what “adequate” testing looks like, and an honest statement of what red-teaming cannot establish.
Read the note Research noteGoverning agentic AI: the open questions
Agents break the loop oversight was built on. The gaps current frameworks leave, and the interim controls we use.
Read the note White paperThe Inception Hack
Prompt injection against enterprise AI: how the attack works, why perimeter thinking misses it, and what actually helps.
Read the paperBring us the question you can’t get answered.
Thirty minutes with a founder. Bring the question the rulebook has not caught up with. If it belongs on the Lab bench we will say so; if it is already answered, we will point you to where.
Price agreed before we start · NDA standard · Lab access is for client partners