References & acknowledgements
The guidance this stack is built with regard to
Balian AI designs its governed stack with regard to published guidance, standards and law. We acknowledge their authors here, with what each source is, how the stack draws on it, and where to read it. Numbered markers elsewhere on this site link to these entries.
Guidance papers
- Guidance paper
Safeguards for Agentic Finance at Runtime (SAFR)
Monetary Authority of Singapore, with financial institutions and FinTechs under the BuildFin.ai initiative · 3 July 2026
A white paper setting out an industry reference approach for verifying an AI agent's actions at the point of decision: agent identity, a controls repository, a disposition engine and a tamper-evident audit log. It is guidance for financial services, not a regulation or a certification scheme.
In this stack. The runtime pattern every action in this stack follows: an agent proposes, a deterministic interceptor decides (allow, observe, escalate or deny) against the agent's mandate, and the decision is written to a hash-chained ledger. We apply the approach to legal practice and human services; we make no claim of MAS endorsement.
Frameworks
- Framework
Model AI Governance Framework for Agentic AI
Infocomm Media Development Authority (IMDA), Singapore · January 2026; updated 20 May 2026
Singapore's voluntary framework for deploying AI agents responsibly: bounding what agents may do, keeping humans meaningfully accountable, technical controls across the agent's life, and end-user responsibility.
In this stack. Least agency and human accountability: agents act only within a written mandate, consequential steps wait for a named person, and every action is attributable to a person, a mandate and a rule.
- Framework
Artificial Intelligence Risk Management Framework (AI RMF 1.0)
US National Institute of Standards and Technology · January 2023
A voluntary framework for managing the risks of AI systems: govern, map, measure and manage.
In this stack. Measurement before use: benchmarks and explanation checks gate every scoring and generative model.
Supervisory guideliness
- Supervisory guidelines
Guidelines on Artificial Intelligence Risk Management
Monetary Authority of Singapore · October 2026
MAS's supervisory expectations for financial institutions on AI governance, risk assessment and the life-cycle management of AI systems.
In this stack. Model governance: every model is benchmarked on a gold set and approved before any request can use it, with approvals recorded and withdrawable.
Laws
- Law
Personal Data Protection Act 2012 (Singapore)
Personal Data Protection Commission, Singapore · In force; as amended
Singapore's general law on the collection, use and disclosure of personal data.
In this stack. Personal identifiers are tokenised before indexing or reaching a model, consent is checked on every use where it is the basis for processing, and records stay inside the organisation's environment.
- Law
Personal Data Protection Act 2010 (Malaysia)
Department of Personal Data Protection, Malaysia · In force; as amended
Malaysia's law on the processing of personal data in commercial transactions.
In this stack. As for Singapore: tokenisation of identifiers (including MyKad numbers and Malaysian addresses), consent checks and data kept in the organisation's own environment.
Standards
- Standard
ISO/IEC 42001:2023, Artificial intelligence: Management system
ISO and IEC · 2023
The international standard for establishing, operating and improving an AI management system.
In this stack. The operating model around the stack: documented decisions, change control for mandates and models, and an audit trail an organisation can show. The stack is not certified against the standard.
A note on these sources. Being informed by a guidance paper, framework or standard is not certification, endorsement or regulatory approval, and none of these bodies has reviewed this stack. SAFR in particular is a guidance paper for financial services; Balian AI applies its runtime approach to legal practice and human services. Nothing on this site is legal advice. Links go to the publishers’ own pages, which may change.
The Balian Papers set out our thinking at greater length, with their own citations.