Guidelines by field
AI guidelines in government and the public sector what the field's own bodies ask
Every major government framework, whatever its politics, asks for human oversight, transparency, appeals and testing. Governments differ most on how much to regulate and who should do it, and those differences sharpened in 2025 and 2026.
Last reviewed 27 October 2026. Every claim links to its source.
At a glance
Respect rights, dignity and the rule of law
The common starting point of the OECD, UNESCO and the Council of Europe.
Say when people are dealing with AI
Transparency and explainability, including telling people they are not talking to a human.
Keep meaningful human control
People keep the final say.
Let people challenge decisions
US federal agencies must offer remedies or appeals for high-impact AI.
Test, monitor and stop what fails
Minimum practices before and after deployment.
Who has spoken
- Binding law, regulation, court ruling or enforceable professional rule
- Voluntary guidance, principles or a code an organization chooses to follow
Voluntary 2019; updated May 2024
OECD
AI Principles
Five values-based principles adopted by 47 governments: human rights, transparency, robustness, accountability, and inclusive growth and sustainability.
Voluntary Nov 2021
UNESCO
Recommendation on the Ethics of AI
Adopted by 193 member states, with human oversight and environmental sustainability among its core values.
Binding Opened Sep 2024; EU ratified May 2026
Council of Europe
Framework Convention on AI, Human Rights, Democracy and the Rule of Law
The first international AI treaty, binding on states that ratify it. Signatories include the US, UK and EU.
Voluntary 2023; 2025
G7 Hiroshima AI Process
International Code of Conduct and reporting framework
A code for advanced AI developers, with public reports through the OECD.
Binding 2025
United States federal government
AI Action Plan; OMB memos M-25-21 and M-25-22; executive orders
The Action Plan calls for removing barriers and for procurement of models it calls truth-seeking and ideologically neutral. The OMB memos, which bind agencies, keep minimum risk practices for high-impact AI, including human oversight and appeals.
Voluntary Feb 2025
United Kingdom
AI Playbook for the UK Government
Ten principles for civil servants. The UK has no general AI statute.
Binding 2023; Sep 2025
China
Interim Measures for Generative AI; labeling measures
Rules for generative AI services and mandatory visible and hidden labels on AI-generated content.
You have meaningful human control at the right stages.
Binding and voluntary
Binding
The EU AI Act (older high-risk systems already used by public authorities must comply by August 2030); the Council of Europe Convention for states that ratify it; OMB memos for US federal agencies; China's measures for providers in China.
Voluntary
The OECD and UNESCO recommendations, the G7 code and reporting, and the UK Playbook.
What changed in 2025 and 2026
- US federal policy shifted from risk management toward acceleration, while keeping minimum practices for high-impact AI, and a December 2025 order targeted state AI laws with a carve-out for child safety.
- The EU delayed its high-risk rules and ratified the Council of Europe treaty in May 2026.
- China's labeling regime took effect in September 2025.
Where people still disagree
Positions differ sharply. The US administration frames its approach as protecting free speech and innovation from ideological bias and regulatory patchworks; critics, including some states and civil-society groups, argue it weakens protections against discrimination. The EU frames its rules as protecting rights; critics there argue they slow innovation, which drove the delays. The primary documents are linked here so readers can weigh them.
Questions to ask yourself
- If this system makes a decision about a member of the public, how do they find out, and how do they appeal?
- Was the system tested on the people it will serve, and who monitors it after launch?
- Could I explain the system's role in plain words at a public meeting?
How this lines up with my everyday guidelines
| My everyday guidelines | The field's nearest rule |
|---|---|
1. Maintain your expertise Use AI to enhance your judgment, not replace it. | Keep meaningful human control (UK Playbook; UNESCO). |
4. Use for good Make sure uses align with your values and professional ethics. | Respect human rights, dignity and the rule of law (OECD; Council of Europe). |
Not part of my everyday guidelines, which are written for one person. | People affected must be able to challenge decisions and seek remedies (Council of Europe; OMB M-25-21). |