MAS-ready AI pre-implementation review pack

No product found does the whole job: map a project to MAS risk dimensions, use that to set how deep the review goes, run the tests, capture agent logs and sign-offs, and export the pack an independent reviewer needs. MAS's final guidelines take effect on 7 October 2027. Secondary summaries say high-risk AI needs formal independent validation before deployment and later re-validation, which creates demand before that date. A new entrant can win by wrapping the free Veritas, AI Verify and Moonshot tests instead of competing with them, adding SAFR-style agent logs and Singapore-specific checks such as whether test data is locally representative, and exporting in a format reviewers accept. Main threats: ValidMind or Credo AI adding a MAS module, KPMG or Resaro building tooling in-house, banks treating this as a feature of existing model-risk tools, and guidelines that allow tailored approaches rather than a fixed pack format. The reviewer edition is harder because the Big Four usually build their own tools; boutique reviewers are the more realistic buyers.

• MAS Veritas, AI Verify and Project Moonshot (IMDA / AI Verify Foundation)

  Free, government-backed testing toolkits: Veritas tests against MAS's fairness and ethics principles (FEAT), AI Verify runs governance tests, Moonshot red-teams language models. Being aligned with each other. This is the free substitute for the 'run the tests' part.

• Resaro

  Singapore AI assurance firm. Premier member of the AI Verify Foundation and a third-party tester; contributed to AI Verify 2.0 and Moonshot; co-ran a finance event with MAS. Closest local rival, as both a vendor and a reviewer.

• KPMG Trusted AI Assurance / Trusted AI Centre of Excellence

  Launched in Singapore in May 2026, backed by the Economic Development Board. Evidence-based assessment of AI systems, with financial services as a target sector. This is the reviewer side; likely to build its own tools.

• Protiviti Singapore

  Advisory service built specifically around the MAS AI Risk Management guidelines.

• Akka Agentic AI Platform

  Append-only, replayable agent logs and signed evidence bundles for financial services, explicitly tied to MAS's SAFR agent framework. Overlaps with the agent-log part.

• ValidMind

  US model-validation and documentation platform for banks with open-source test suites. Built for US, UK and EU rules; no MAS-specific product found. Could add one.

• Credo AI

  AI governance platform with policy packs for the EU AI Act, NIST and ISO 42001. No MAS pack found, but easy for them to add.

• Intelligible

  Small Singapore startup founded in 2024 doing automated fairness, robustness and security testing plus a model registry. Very little funding; no bank customers found.

• Pecuniya

  Markets help building AI governance programmes aligned with MAS FEAT and the AI risk guidelines. Looks like content and advisory, not a full workbench.

• Samta.ai

  Claims mapping to MAS Technology Risk and FEAT controls. Marketing claims; not verified.

• corvair.ai

  Has a MAS AIRG guide page for Singapore banks. A direct search found nothing more, so its product is unknown.

• Sasana / GuardClaw / ProofRail

  Open-source and early-stage tools for hash-chained, tamper-evident agent logs. The logging layer is becoming a commodity.