Every reviewed AI governance resource in one place — reports, standards, guidance and regulation.
A UK industry temperature check on barriers and enablers to AI assurance. It synthesises stakeholder engagements and proposes sector-specific interventions for HR and recruitment, finance, and connected and automated vehicles.
This guidance sets out 15 security principles for the design, development, deployment, operation and end of life of machine-learning systems. It gives goals, rationale and implementation options for addressing ML-specific vulnerabilities alongside established cyber security practice.
UK government guidance for organisations procuring and deploying AI in recruitment. It maps assurance mechanisms across procurement, piloting and live operation, with examples of recruitment risks and mitigations.
A voluntary set of guidelines for pre-deployment testing of LLM-based applications. It provides methods for assessing five safety and reliability risks through benchmarks, red teaming, output testing and component testing.
This landscape report examines how concentrated AI industry power shapes markets, public institutions, work and infrastructure. It offers an advocacy and policy roadmap focused on worker organising, restrictions on harmful uses, lifecycle regulation and independent oversight.
This guidance presents a Job Impact Assessment Tool and stakeholder-specific recommendations for improving AI’s effects on jobs, job quality and labor markets. It is intended for AI creators and users, policymakers, labor organizations and workers.
This one-page summary describes the CSET AI Harm Framework, a modular structure for defining and categorising harms linked to AI systems. It outlines harm classifications, customisation options, and an example taxonomy for the AI Incident Database.
A three-page summary presenting an analytic approach for identifying the assumptions behind frontier AI governance proposals. It applies three questions to five U.S.-centric proposals and draws policy considerations from the comparison.
This CSET Data Brief evaluates four frameworks for classifying AI systems along policy-relevant characteristics. It reports survey experiments testing the consistency and accuracy of classifications by public users.
A framework proposing nine dimensions for addressing generative AI risks while supporting innovation. It sets out policy and industry considerations for accountability, data, development, incident reporting, assurance, security, provenance, alignment research and public good.
A one-page CSET briefing that summarises three principles for US AI governance: understanding risks and harms, building AI literacy, and preserving adaptable policy. It argues for flexible, iterative governance as AI evolves.
A two-page CSET summary of a harmonized framework for governing, managing, and protecting technology while integrating AI. It outlines 258 recommendations, an AI score, and 34 topic areas across five categories.
A generated AI Verify summary report for a binary credit-risk classification model. It combines process-check results across 11 AI ethics principles with fairness, robustness and explainability test outputs.
A three-page set of policy proposals for UK AI governance. It recommends government expertise, AI assurance and risk-management measures, and international cooperation on safe and responsible AI.
This issue brief compares Anthropic’s Responsible Scaling Policy with UK DSIT guidance on responsible capability scaling. It proposes more verifiable risk thresholds, lifecycle monitoring, government reporting, pause planning and earlier external verification.
This two-page summary presents six findings and 11 recommendations from workshops examining lessons from standards in other industries for AI governance.
A two-page CSET summary of a July 2023 roundtable on anticipated risks from frontier AI systems. It outlines dangerous capabilities of concern and recommendations for government expertise, transparency, third-party testing, and crisis management.
A briefing paper proposing a lifecycle-based framework for identifying AI risk pathways and selecting UK domestic and global policy responses. It sets out policy levers for visibility, best practices, incentives and regulation.
This report presents an open-source intelligence method for detecting AI scheming-related incidents through publicly shared transcripts. It details a collection, screening, scoring and deduplication pipeline and reports findings from X posts collected between October 2025 and March 2026.
This report sets out how frontier AI companies could apply the Three Lines Model to separate risk ownership, oversight and audit. It offers a proposed organisational mapping, a preliminary gap analysis, risk-culture measures and recommendations for companies and government.
This paper examines compute as a lever for governing AI development and deployment. It outlines compute’s distinctive properties, policy mechanisms for visibility, allocation and enforcement, and guardrails against privacy, economic and centralisation risks.
This report argues that frontier AI safety frameworks need organisation-wide risk governance alongside technical thresholds and mitigations. It provides a component model, an indicative comparison of four company frameworks, and recommendations for companies and governments.
A policy brief examining whether Article 50 of the EU AI Act requires AI-agent actions to be labelled. It proposes metadata, verification and standardisation measures for a forthcoming Code of Practice.
This policy brief interprets Measure 7.1, point (4) of the EU GPAI Code of Practice on model specifications. It proposes a four-part rubric covering intended behaviour, principle detail, conflict management and system prompts.
Curated Library of AI Governance Resources
