The frameworks that apply to you

A voluntary framework organised around four functions: govern, map, measure, and manage. Accompanied by a Generative AI Profile addressing risks specific to foundation models.

General cybersecurity risk framework covering govern, identify, protect, detect, respond and recover. Not AI-specific, but the control baseline most AI security expectations assume is already in place.

A management system standard for artificial intelligence, structured like ISO 27001. The only AI framework currently supporting independent third-party certification.
Certification proves you have a system for managing AI risk. It does not prove any individual AI system behaves as intended.

A practical standard for assessing the impacts of AI systems on people, organisations and society. It provides a structured methodology for identifying, evaluating and documenting potential harms, benefits and affected stakeholders throughout the AI lifecycle.

Guidance on AI risk management, applying the ISO 31000 risk process to AI-specific concerns. Frequently used alongside ISO/IEC 42001 as the risk methodology beneath the management system.

A certifiable standard built specifically for AI agents, spanning six risk pillars: safety, security, reliability, accountability, data and privacy, and society. Certification follows independent third-party audit, with quarterly adversarial retesting and crosswalks to ISO 42001, NIST AI RMF, MITRE ATLAS and OWASP.

An adversarial threat landscape knowledge base for AI systems, structured like MITRE ATT&CK. Catalogues real-world tactics and techniques used against machine learning and generative AI systems.

A community-maintained list of the most critical security risks in large language model applications, covering prompt injection, insecure output handling, training data poisoning, excessive agency and related failure modes.

Published October 2025, this is now the primary Australian Government guidance for responsible AI adoption. It condenses the ten VAISS guardrails into six essential practices: decide accountability, understand impacts, measure and manage risks, share information, test and monitor, and maintain human control. Available in Foundations and Implementation editions.

Published 5 September 2024. Ten voluntary guardrails applying across the AI supply chain, covering accountability, risk management, data governance, testing, human control, transparency, stakeholder engagement and supply chain requirements.

Announced 2 December 2025. Sets Australia's regulatory posture: govern AI through existing technology-neutral laws and sector regulators, supported by voluntary guidance and an advisory AI Safety Institute, rather than a standalone AI statute. The September 2024 proposals paper on mandatory guardrails for high-risk AI was not legislated.

Existing privacy law applies in full to AI systems processing personal information. OAIC has issued specific guidance on commercially available AI products and on developing and training generative AI models.

Issued June 2024. Recommendations and best practices for organisations procuring, implementing and using AI, covering AI strategy and governance, risk assessment and human oversight, customisation and management, and communication with stakeholders.

Practical checklist for building an internal generative AI policy, covering scope of permissible use, protection of personal data, lawful and ethical use, data security, and consequences of policy violation.

A series of circulars and papers to authorised institutions, including high-level principles on AI, consumer protection in respect of big data analytics and AI, and guidance on generative AI in financial services. In March 2026 the HKMA joined the SFC, Insurance Authority and MPFA in launching GenA.I. Sandbox++.

Operational guidance for technology developers, service providers and users of generative AI. Covers scope and limitations of application, governance principles, and technical risks including data leakage, model bias and output error.

Framework setting out dimensions of trusted generative AI including accountability, data, trusted development and deployment, incident reporting, testing and assurance, security, content provenance and safety.

Launched January 2026 and described by IMDA as the first governance framework addressing agentic AI specifically, building on the earlier traditional and generative AI frameworks.

An open-source testing framework and software toolkit that allows organisations to run technical tests against AI systems and produce standardised reports on governance principles.

Cybersecurity technology: basic security requirements for generative artificial intelligence services. Released 25 April 2025, effective 1 November 2025. Covers training data security, model security and safety measures. Targets services with public opinion or social mobilisation attributes, and supports filing, testing and evaluation.

Published July 2026. Addresses the security discipline expected of organisations deploying and operating AI agents, covering permissions, identity, monitoring and containment of agent actions within the deploying organisation's infrastructure.

Measures for the administration of labelling of AI-generated content, effective 1 September 2025, with the accompanying mandatory national standard specifying visible labels and embedded metadata identifiers for text, image, audio, video and virtual scenes.

Risk-tiered regulation of AI systems placed on or affecting the EU market, with prohibited practices, high-risk system obligations, general-purpose AI model duties, and transparency requirements. Obligations phase in on a staged timetable.

Requires that people are informed when they interact with an AI system, that synthetic content is machine-readably marked, that deepfakes and AI-generated public interest text are disclosed, and that emotion recognition and biometric categorisation use is notified. Applied from 2 August 2026.

No cross-cutting AI statute. Five cross-sectoral principles are applied by existing regulators including the ICO, FCA, CMA, Ofcom and MHRA within their own remits.