Curated · Practitioner-rated
30 frameworks, reports, guidelines, and tools — curated for AI governance practitioners.
Best practice guidance from Partnership on AI for developing AI systems responsibly in safety-critical contexts. Covers risk assessment, testing, monitoring, and incident response for high-stakes AI deployments. Developed with input from leading AI companies, academics, and civil society.
A comprehensive framework from IEEE for aligning AI and autonomous systems with human values and ethical principles. Covers transparency, accountability, data agency, effectiveness, and value alignment. Serves as a basis for IEEE's suite of AI ethics standards.
A knowledge base of adversarial machine learning tactics and techniques based on real-world observations. Modelled after the MITRE ATT&CK framework, covering attack stages from reconnaissance to impact specific to AI/ML systems. Essential resource for AI red-teaming and threat modelling.
Multi-stakeholder resources on responsible AI development covering safety, fairness, accountability, and transparency. Includes case studies, best practice guides, and research from across industry, academia, and civil society. Developed collaboratively by PAI's diverse membership.
Proposes that every dataset be accompanied by a datasheet documenting motivation, composition, collection process, and recommended uses. Analogous to datasheets in electronics manufacturing, aimed at increasing transparency and accountability in ML. Widely adopted companion to model cards.
A set of ten ethical principles for AI development collaboratively developed by citizens, experts, and stakeholders. Covers well-being, autonomy, justice, privacy, knowledge, democracy, and sustainability. Developed through a deliberative democratic process involving thousands of participants.
Comprehensive annual report tracking AI developments across research, technical performance, economy, education, policy, and ethics. Provides data-driven insights on the state of AI globally with hundreds of charts and tables. Widely referenced by policymakers, researchers, and practitioners.
Proceedings of the ACM Conference on Fairness, Accountability, and Transparency, the premier academic venue for research on algorithmic fairness and accountability. Covers technical, legal, and sociotechnical approaches to responsible AI. Annual conference bringing together researchers and practitioners.
Research paper introducing Constitutional AI (CAI), a method for training AI systems to be helpful, harmless, and honest using a set of principles. Proposes using AI feedback rather than purely human feedback to scale safety training. Foundational work for aligning large language models.
The first intergovernmental AI standard, providing five principles for responsible AI and five recommendations for policy. Endorsed by G20 and incorporated into many national AI strategies. Serves as a foundation for international AI governance discussions.
International standard providing guidance on how organisations can manage risk specific to AI. Extends ISO 31000 risk management principles to the AI context, covering the full AI system lifecycle. Complements ISO 42001 and provides detailed risk management processes.
An accessible introduction to algorithmic accountability covering key concepts, case studies of algorithmic harms, and policy recommendations. Examines how algorithms affect criminal justice, hiring, lending, and healthcare. Essential primer for policymakers and advocates.
Policy analysis and guidance from Brookings on managing risks from generative AI in enterprise settings. Covers governance structures, vendor risk management, and regulatory compliance strategies. Informed by expert interviews and analysis of current practices.
Official guidance from the European Commission on complying with the EU AI Act. Explains the risk classification system, key obligations for providers and deployers, and implementation timelines. Essential starting point for any business seeking to understand EU AI Act compliance requirements.
Practical guidance from the UK data protection regulator on how to apply data protection law when developing and deploying AI. Covers lawful basis for processing, fairness, transparency, and data minimisation in AI contexts. Includes an AI and data protection risk toolkit.
Companion resource to the AI RMF providing suggested actions, references, and relationships for each subcategory of the framework. Helps practitioners implement the AI RMF in their organisations. Available as an interactive online resource and downloadable document.
Seminal paper proposing model cards as short documents accompanying AI models to communicate their intended uses, performance characteristics, and ethical considerations. Widely adopted as an industry best practice for AI transparency. The standard approach for documenting AI models.
The first global normative instrument on AI ethics adopted by 193 member states. Covers human rights, fairness, transparency, accountability, and sustainability in AI development and deployment. Includes a readiness assessment methodology for member states to evaluate their AI ethics governance.
Research reports from CSET covering AI policy, national security implications of AI, and AI governance. Provides rigorous analysis of AI developments for US policymakers and researchers. Covers topics including AI talent, compute governance, and military AI.
A profile of the AI RMF specifically addressing generative AI risks and trustworthiness challenges. Covers 12 generative AI-specific risks including hallucination, data poisoning, and misuse for disinformation. Provides practical guidance for managing these risks within the broader AI RMF.
A conceptual framework for securing AI systems throughout their development and deployment lifecycle. Covers six core elements including expanding strong security foundations, extending detection and response to AI pipelines, and automating defences. Intended to complement existing cybersecurity frameworks.
A public repository of AI incident reports documenting cases where AI systems caused harm or behaved unexpectedly. Searchable by incident type, sector, technology, and harm. Valuable for risk assessment, AI audits, and understanding real-world AI failure modes.
Briefing papers from the WEF AI Governance Alliance covering responsible AI deployment across industries. Brings together industry, government, and civil society leaders to develop actionable AI governance guidance. Covers topics including generative AI governance, AI in healthcare, and financial services.
The official text of the EU Artificial Intelligence Act, the world's first comprehensive AI regulation. Establishes a risk-based classification system for AI systems with tiered obligations for providers and deployers. Essential reading for any organisation placing AI systems on the EU market.
The definitive voluntary framework for managing AI risks, organised around four functions: Govern, Map, Measure, and Manage. Provides practical guidance for integrating AI risk management into organisational processes. Widely adopted as a baseline by US federal agencies and private sector organisations.
Annual reports from the AI Now Institute examining the social implications of AI including accountability, bias, and surveillance. Provides critical analysis of AI industry practices and policy developments. Essential reading for understanding the societal dimensions of AI governance.
The first certifiable international standard for AI management systems, specifying requirements for responsible AI development and use. Helps organisations demonstrate accountability and trustworthiness in their AI systems. Aligned with ISO 9001 and ISO 27001 structures for integration into existing management systems.
Documents the ten most critical security vulnerabilities in LLM-based applications, including prompt injection, insecure output handling, and training data poisoning. Provides prevention and mitigation guidance for each vulnerability. Essential reading for teams building or securing LLM-powered products.
Policy analysis and advocacy resources on AI governance with a focus on existential and catastrophic risk from advanced AI. Covers frontier AI regulation, international governance frameworks, and technical AI safety. Includes policy papers, open letters, and expert briefings.
A maturity model helping organisations assess and improve their responsible AI capabilities across six dimensions: strategy, culture, process, data, model, and infrastructure. Provides a roadmap for organisations at different stages of AI governance maturity. Includes certification pathways.