AIRR Rapid Access award
20,000 GPU hours on Isambard-AI
Awarded through the UKRI AI Research Resource (AIRR) Rapid Access programme.
Our research connects private AI engineering with the practical demands of regulated work. Useful progress includes finding where a model, a retrieval method or an evidence process should not be trusted.
With research on Isambard-AI, part of the UK’s AI Research Resource, operated by the University of Bristol, and a separate AIRR Cloud pilot award.
A score has value when its test population, conditions, failures and limits are understood.
Can each important statement be traced to the exact admitted source, and do missing, changed or contradictory sources remain visible?
Which facts, conditions, negations and source links survive repeated context compression, interruption and restart?
Which model, retrieval and context configuration performs the defined work within its quality, privacy and operating constraints?
Can an authorised reviewer reconstruct the basis, decisions, changes and limitations without relying on an unexplained model response?
Comparisons set in advance, and negative results kept, stop a persuasive demonstration becoming a claim the evidence can’t support.
Question, sources, cases, baseline and acceptance criteria.
Useful evidence, failures, uncertainty and complete cost.
Missing sources, contradictions, hostile input and recovery.
Inputs, results, failed attempts, corrections and limitations.
Improve, retain, narrow, investigate or stop.
Model agreement is not independent ground truth. Context compression is not perfect memory. A successful source check is not a professional judgement. Each needs the right evaluation and evidence. What’s built and working is listed in our trust centre.
Each is credited in its own required wording.
20,000 GPU hours on Isambard-AI
Awarded through the UKRI AI Research Resource (AIRR) Rapid Access programme.
This project received support through an AI Research Resource (AIRR) Cloud pilot award, funded by the UK Government and delivered in partnership with the STFC Hartree Centre.
These public sources inform the context and the evaluation approach. None is an endorsement, certification or assessment of Aprentiz.
Bank of England and FCA, 2024
75% of responding firms reported using AI; 46% reported only partial understanding of the technologies they use. The survey also examines model complexity and third-party dependencies.
Financial Conduct Authority
The FCA’s public material describes safe and responsible adoption, development and testing of AI in UK financial markets.
NIST
A voluntary framework for considering AI risks across design, development, use and evaluation. Aprentiz makes no claim of certification or complete conformance.
Isambard-AI
The official Isambard-AI publication citation, facility statement and separate Cloud pilot acknowledgement are kept on our acknowledgements page.