MINERVA Studio

0 reviews

MINERVA keeps one record of every model, vendor AI tool and AI agent a bank runs, with every change and sign-off on it, and hands an examiner an evidence packet they can check without us. Built for OSFI Guideline E-23, which takes effect on 1 May 2027. Banks and insurers regulated by OSFI must keep a model inventory, rate every model's risk, have it reviewed by someone who did not build it, monitor it, and prove all of that to an examiner. Most do it in spreadsheets and email. MINERVA turns that into one record: - Model inventory in E-23's own fields, imported from the spreadsheet you already keep. - Approval controls that check before a move: the person who built a model cannot approve it, a review needs its evidence, and an open high-severity finding blocks approval. - Monitoring and escalation: watched figures against recorded lines; a breach becomes a proposal that a named person decides. - Vendor and AI agent governance: third-party models under OSFI B-10, and AI agents governed by policy as code that a person approves. - Audit trail with outside time-stamps, and an evidence packet the examiner verifies offline with one command. - Rule packs on the same record: NIST AI RMF, ISO/IEC 42001, the EU AI Act, SR 26-2 and PRA SS1/23. A named person always decides; the software proposes and records. For model risk, validation and compliance teams at federally regulated financial institutions in Canada, and the consultancies that run E-23 readiness work for them.