Skip to content
Xplore
← Cambridge Tech Week hub
For universities and research groups

Universities own the method. We own the instrument.

A verdict on an AI agent is worth nothing if it comes from the agent's vendor — or from ours. So the methodology, the calibration and the signature database are being placed with UK universities, coordinated by Edinburgh Napier. We bring the engine, the schema and the traces. You bring the science.

What a partner gets
  • 1Data. Traces from the public testbed (96 external agents, nine cases) and, under agreement, anonymised traces from production — including runs on runtimes we do not control.
  • 2An instrument to run against. The deterministic verification engine (19 checks), the open trace schema, and the capture bridges.
  • 3Authorship. Co-authorship on the method; a seat at the October workshop and on the bids that follow.
Where it goes

TRL 5 in 2026–27; a consortium-authored signature database in 2027–28; an accreditation-ready test facility — re-runnable verdicts admissible to regulators and insurers — by 2029. The science is grant-funded through joint bids; Xplore’s own round funds sales, not research. Partners are not asked to invest in the company.

Where the programme stands
  • doneJul 2026 TRL 4 assessment of the platform by Edinburgh Napier University. 5,078 scored cases. Letter ref. IV-XPLORE-2026.
  • 2026 TRACE-AI — a contradiction-aware provenance benchmark for agentic pipelines. Bid led by Edinburgh Napier with UCL; Xplore as industry partner. Decision pending.
  • in motionAug 2026 After a ministerial meeting, Edinburgh Napier was asked to coordinate a Scottish consortium on agent assurance. Concept note circulated.
  • nextOct 2026 Scoping workshop: strand leads, position statement, the next joint bids, an accreditation-landscape study.
Research topics · open, to be claimed in October
Signature taxonomyClassifying malicious and incorrect agent behaviours; who authors and retires a signature.
Capture across runtimesRecording an agent at the gateway, without its code; what each capture level loses.
Passport metrologyCalibration and error bounds of a run passport.
Sampling vs every runWhat changes when every run is checked at flat cost, not a sample.
Provenance benchmarksContradiction-aware provenance for agentic pipelines (the TRACE-AI line).
Formal controlSpecifying what an agent may read, write and call; multi-agent coordination.
Clinical validityDoes a testbed verdict transfer to a regulated clinical workflow?
Economics of assuranceWhat an independent verdict is worth to a buyer, a platform, an insurer.
Human factorsHow a reviewer reads a passport and acts on a located risk.
Where a joint bid can go · indicative
  • Innovate UK · KTPUniversity-led, an associate placed in the company; SME rate up to 67 %; yearly rounds.
  • Innovate UK · Frontier AI Phase 2Collaborative consortia, £5–10M, 24–32 months; academic share 20–30 %.
  • EPSRC · Prosperity PartnershipsLong-term industry–academic agenda; or responsive mode for one sharp question.
  • EPSRC · iCASE studentshipsIndustrial PhDs co-defined with the company, on testbed and production traces.
  • Responsible AI UKKeystone and impact-accelerator calls on trustworthy, accountable AI.
  • Interface · Scottish EnterpriseInnovation vouchers for a first joint project; SMART: Scotland and collaborative R&D after.
  • The Data LabScotland's data and AI innovation centre: collaborative projects, MSc placements.
  • NIHR i4iFor the clinical strand, with an NHS partner, building on the medtech deployment.
  • Horizon Europe · Cluster 4Consortium-scale bids on trustworthy AI; the UK is an associated country.

TRACE-AI, the first joint bid (Edinburgh Napier, UCL, Xplore), is submitted and awaiting decision. Schemes and rounds change; we match the scheme to the strand and the partner at the workshop.

Pick one

Xplore Intelligence Ltd · Edinburgh · hello@xploreintelligence.co.uk · Stand: Innovation Alley, Tue 15 – Wed 16 Sep