AI governance is the set of rules, institutions and practices that steer how artificial intelligence is developed, deployed and used. It operates on several levels at once: inside companies, through technical standards, in national and state law, and in international agreements and forums. No single authority governs AI; the result is a patchwork that keeps changing.
Companies and standards
Much governance happens before any law applies. AI developers publish usage policies, model specs, system cards and frontier safety frameworks such as Responsible Scaling Policies. Voluntary standards give them a common language: the US National Institute of Standards and Technology published its AI Risk Management Framework in January 2023, and ISO/IEC 42001, the first certifiable management-system standard for AI, appeared in December 2023.
National and regional law
The EU AI Act, adopted in 2024, is the most comprehensive law: it sorts AI uses by risk, bans some practices and sets duties for general-purpose models, though in 2026 the EU agreed to postpone most of its high-risk rules to 2027 and 2028. Supporters expect its influence abroad to follow the so-called Brussels effect; critics doubt it. The United States has no comprehensive federal AI law. It relies on executive action, such as the White House’s AI Action Plan of July 2025 and a December 2025 executive order directing the Justice Department to challenge state AI laws, on existing agencies and on state laws, such as California’s SB 53 on transparency for frontier models (2025) – which in turn fuels a debate over federal preemption. China regulates through targeted rules on recommendation algorithms, deep synthesis and generative AI, with a registry of algorithms and labeling duties.
International forums
The OECD AI Principles (2019), the G7’s Hiroshima Process (2023) and the Council of Europe’s Framework Convention on AI (opened for signature in 2024) set shared expectations. A series of summits – Bletchley Park (2023), Seoul (2024), Paris (2025), New Delhi (2026) – brought governments and companies together. In August 2025 the UN General Assembly created an Independent International Scientific Panel on AI and a Global Dialogue on AI Governance; the panel presented its preliminary report and the Dialogue held its first annual session in Geneva in July 2026, with the next session planned for New York in May 2027.
The big choices
Behind the institutions lie a few recurring choices: licensing before deployment or liability after harm; the precautionary principle or permissionless innovation; control of inputs such as chips (compute governance) or rules for uses; and how to keep rule-making from being captured by the largest firms (regulatory capture). Tools such as regulatory sandboxes try to combine experimentation with oversight.
Why it matters
Governance decides who may build what, who bears the risks and who shares the benefits. aiciety’s view is that many overlapping centers of rule-making – companies, courts, states, standards bodies and international forums – learn faster than a single master plan, provided the rules focus on outcomes and keep markets open.
Written with Claude, made by Anthropic, which publishes a Responsible Scaling Policy and system cards, supported California’s SB 53, has advocated chip export controls, opposed a blanket federal moratorium on state AI laws, is in a legal dispute with the US Defense Department and is subject to many of the rules described here.
Related
EU AI Act · AI Safety Institute · Compute Governance · Preemption · Regulatory Sandbox · AI Ethics · AI Safety · Brussels Effect
Sources: NIST: AI Risk Management Framework (January 26, 2023) · ISO: ISO/IEC 42001:2023, AI management systems (December 2023) · The White House: Winning the Race: America’s AI Action Plan (July 23, 2025) · Morrison Foerster: Executive order on state AI laws (December 2025) · White & Case: EU agrees Digital Omnibus deal to simplify AI rules (May 2026) · Office of Governor Gavin Newsom: Governor Newsom signs SB 53 (September 29, 2025) · Council of Europe: The Framework Convention on Artificial Intelligence · United Nations: Global Dialogue on AI Governance, FAQ (2026)
Leave a Reply