Practical requirements extracted from policy instruments, with the source article, the actors they bind, evidence examples and original framework mappings. Legal requirements are marked; everything else is voluntary guidance.
86 of 117 obligations match · page 4 of 4 · Download these as CSV
Singapore Model AI Governance Framework · Second edition, Part on human involvement in AI-augmented decision-making
Using a risk-impact matrix (probability and severity of harm), organisations choose human-in-the-loop, human-over-the-loop or human-out-of-the-loop designs and document the rationale.
UK AI regulation framework · Principle 5, Part 3
Affected people should be able to contest harmful AI decisions or outcomes and obtain redress, through existing complaint routes and regulators.
Australian Voluntary AI Safety Standard · Guardrails 4, 5 and 6
Test AI models and systems before deployment and monitor them in operation; enable meaningful human control and intervention; and inform end users about AI-enabled decisions, interactions with AI and AI-generated content.
NIST AI RMF · MAP function
Map establishes the context: intended purposes, users, deployment settings, legal requirements, risk categorisation, benefits and costs, and impacts on individuals, groups, communities and society.
AI Basic Act · Article 35
An AI business operator that provides high-impact AI, or a product or service using it, is to make efforts to assess in advance the impact the AI may have on people's fundamental rights. The provision is drafted as an endeavour duty rather than a hard requirement, but public bodies procuring high-impact AI are directed to give preference to products that have undergone such an assessment, and the ministry may set assessment methods.
UK AI regulation framework · Principle 1, Part 3
Regulators are asked to ensure AI systems function in a robust, secure and safe way, with risks continually identified, assessed and managed. In practice this is enforced through existing safety, security and data-protection law rather than a new duty.
NIST AI RMF · MANAGE function
Manage allocates resources to mapped and measured risks, plans responses including decommissioning, manages third-party risks, and documents post-deployment monitoring, incident response and communication.
NIST AI RMF · MEASURE function
Measure covers selecting metrics and test methods, evaluating validity, safety, security, resilience, explainability, privacy, fairness and bias, and monitoring these over time, including through independent review and red-teaming for generative AI.
UK AI regulation framework · Principle 2, Part 3
Organisations should communicate when and how AI is used and provide explanations proportionate to the risk, so that people can understand decisions affecting them. For personal data, UK GDPR transparency and automated decision-making rights make this binding in practice.
Singapore Model AI Governance Framework · Generative AI framework, dimensions on incident reporting and content provenance
The generative-AI framework recommends incident-reporting channels and processes for AI harms, and content provenance measures such as digital watermarking and cryptographic provenance so that users can identify AI-generated content.
Australian Voluntary AI Safety Standard · Guardrails 7, 8 and 9
Establish processes for people impacted by AI to challenge use or outcomes; be transparent with other organisations across the AI supply chain about data, models and systems; and keep and maintain records to allow third parties to assess compliance.
Filters
A single practical requirement pulled out of an instrument and stated on its own: keep a risk management system, log incidents, document training data, provide human oversight, and so on. Each one cites the article or section it comes from so you can check it against the source.
No, and every obligation is labelled either a legal requirement or voluntary guidance. Voluntary items still matter in practice, because procurement questionnaires and auditors ask about them, but only the binding ones carry legal consequence.
Filter by jurisdiction, category, actor, sector or use case. The applicability check asks a short set of questions and returns the duties that may reach you. It is an educational screen, not a legal determination, and it says so.
Because nobody has broken them out yet. Most instruments are recorded at summary level first; obligations are added jurisdiction by jurisdiction. The coverage and open-gaps pages publish exactly what is missing rather than hiding it.
