What Is AI Governance? Definition, Examples and First Steps
AI governance is the roles, rules and records that decide which AI systems your organisation uses and who answers for each one. Definition, a worked example, the laws and standards involved, and a 30-day start plan.

On this page
In short: AI governance is the roles, rules and records that decide which AI systems your organisation uses and who answers for each one. It covers AI agents, automations and applications, whether you build them, buy them or switch them on inside software you already pay for.
If someone asked you tomorrow which AI systems you run and who signed each one off, AI governance is what lets you answer from records instead of from memory.
What is AI governance?
AI governance is how an organisation stays in charge of the AI it uses. In practice it has three working parts:
- People. A named owner for each AI system, a person who approves it before go-live, and a route to the board or trustees.
- Rules. What AI may be used for, what needs approval, what is banned, and which laws and standards apply to each system.
- Records. A register of systems, signed approvals, logs, incidents and training. The records turn the rules into something an auditor can check.
A policy document, an ethics statement or a one-off audit is not AI governance. And your AI supplier can’t do it for you: as the deployer, you keep your own duties.
AI governance in one example
This is a hypothetical example, not a customer story.
A housing association adds an AI agent to its tenant helpdesk. The agent can read tenancy records, answer questions and log repair requests. Three months later, a tenant complains that the agent told them their rent arrears would be written off. Nobody can say who approved the agent, what it was allowed to promise, or which version was running that week.
With AI governance in place, the same story runs differently. The agent is on the AI register with a named owner, the head of tenant services. Before go-live, the owner and the data protection officer signed an approval that listed what the agent may and may not do, including “no statements about arrears or payments”. A guardrail blocks that topic and hands the tenant to a person. When the complaint arrives, the log shows the exact conversation, and the incident goes on the register the same day.
Same technology. The difference is that someone answers for it, and the evidence exists.
Why AI governance matters now
AI use is ahead of oversight. OutSystems’ 2026 State of AI Development report, based on about 1,900 IT leaders, found that 96% of organisations already use AI agents, but only 12% have a central platform to manage them.
Many agentic AI projects are expected to be cancelled. Gartner has warned that over 40% of agentic AI projects will be cancelled by 2027 without proper governance and clear value.
Some duties already apply. Under the EU AI Act, the banned practices in Article 5 and the AI literacy duty in Article 4 have applied since 2 February 2025, and chatbot transparency under Article 50 since 2 August 2026. Two new bans and the end of a labelling grace period arrive on 2 December 2026; The EU AI Act delay is real: here’s what still applies on 2 December 2026 has the full timeline. GDPR applies to any AI system that processes personal data. Fines for prohibited practices reach up to €35m or 7% of worldwide annual turnover (Article 99 of Regulation (EU) 2024/1689, the EU AI Act).
Evidence can’t be back-filled. A regulator, auditor or customer will ask who approved a system and what changed since. If you didn’t record it at the time, you can’t credibly recreate it later.
AI governance vs AI ethics, AI compliance and data governance
| What it asks | Output | Who usually leads | |
|---|---|---|---|
| AI ethics | Should we do this at all? | Principles and values | Leadership, ethics board |
| AI compliance | Do we meet the specific laws that apply? | Legal assessments, conformity records | Legal, compliance |
| Data governance | Is our data accurate, protected and used lawfully? | Data catalogue, data protection impact assessments | Data lead, DPO |
| AI governance | Who decides, under which rules, and can we prove it? | AI register, approvals, logs, incidents | Accountable owners, with a senior sponsor |
AI governance is the operating layer that connects the other three to real systems. Ethics sets the values, compliance names the legal duties, and data governance covers the inputs. AI governance makes sure each AI system has an owner, an approval and a record. For the overlap with data, see AI governance vs data governance: what is the difference and do you need both?
Which laws and standards shape AI governance?
No single law defines AI governance. Most organisations govern against several documents at once.
| Document | Type | What it adds |
|---|---|---|
| EU AI Act, Regulation (EU) 2024/1689 | Law (EU) | Risk categories, banned practices, duties for providers and deployers, AI literacy |
| GDPR, Regulation (EU) 2016/679 | Law (EU) | Lawful basis, data protection impact assessments, rights around automated decisions (Article 22) |
| ISO/IEC 42001:2023 | Certifiable standard | An AI management system: policy, roles, risk assessment, continual improvement |
| NIST AI RMF 1.0 (January 2023) | Voluntary guidance (US) | Four functions: Govern, Map, Measure, Manage |
| OECD AI Principles (2019, updated 2024) | Intergovernmental principles | Shared values adopted by many governments |
| National laws, for example the Texas Responsible AI Governance Act and South Korea’s AI Basic Act | Law (national or state) | Local duties that sit alongside the above |
Sector rules apply on top: financial services, health, education and public bodies each have their own. Check the current text of any law before acting on a single date. The EU AI Act itself was amended in 2026 by the Digital Omnibus on AI. If you are choosing between the two main reference frameworks, read NIST AI RMF vs ISO 42001.
Who is responsible for AI governance?
Accountability sits with the organisation that deploys the AI, not with the model provider. Inside the organisation, the roles usually look like this:

- The board or trustees set the appetite for AI risk and receive regular reports.
- A senior sponsor owns the AI governance programme as a whole.
- An accountable owner for each system. One person, not a team. A team is not an owner.
- Approvers sign off a system before go-live and before material changes. From 2 December 2027, under Article 26 of the EU AI Act, deployers of high-risk systems listed in Annex III must assign human oversight to people with the competence, training and authority to do it.
- Compliance, risk and the data protection officer classify systems and check the legal duties.
- Everyone who uses AI needs enough AI literacy to know what the system does, where it fails and when to escalate.
Five principles behind it
We group the work into five principles. Each one should show up as something you can point to.
| Principle | What it means in practice | The record that proves it |
|---|---|---|
| Accountability | A named person answers for each system | Owner on every row of the AI register |
| Proportionality | Controls match the risk of the use | Risk classification with the reason |
| Human oversight | People approve outcomes that matter | Signed approval gates |
| Transparency | People know when AI is involved and why | Disclosures, model and data records |
| Evidence | Behaviour is checked while the system runs | Logs, monitoring results, incident register |
How to start in 30 days
Five steps, four weeks. You don’t need a finished framework first.
- Week 1: list every AI system. Include agents, automations and applications you built, tools you bought, AI features switched on inside existing software, and tools staff use without approval. That last group is shadow AI.
- Week 1: name one owner per system. If nobody will own it, that is your first finding.
- Week 2: classify each system. Note which laws apply, and flag anything that touches hiring, credit, education, essential services or health.
- Week 3: put an approval gate in front of the riskiest system. A person reviews it and signs, and the signature is kept.
- Week 4: write a one-page acceptable-use policy and record who read it. That record is also your first AI literacy evidence.
When you’re ready for the full structure, our guide to the AI governance framework: the 8 parts and a template maps each part to the EU AI Act, ISO/IEC 42001 and the NIST AI RMF.
How @Human by Planet Crust helps
Planet Crust @Human is the compliance-first, open-source AI platform for organisations that answer to somebody. You build AI agents, automations and applications in one place, with the governance already there.
- Every system is classified against the EU AI Act and ISO/IEC 42001, with a risk register and a sensitive-data register. @Human is ready for GDPR and for any country’s laws and regulations.
- Accountable roles and signed approval gates. AI agents can’t sign a gate: 0 of the 296 tools on a live @Human instance can.
- Runtime guardrails, monitoring and incident tracking, so evidence is recorded as the system runs.
- Self-hosted and sovereign. Your data and models stay with you, you can use any AI provider, and you can always unplug.
@Human gives you the controls and the evidence. Whether you are compliant still depends on how you use them. Pricing is pay what you can, from €1 per month, and not-for-profits get a free build.
Frequently asked questions
What is AI governance in simple terms?
It is how an organisation decides which AI it uses, who answers for each system, and how it proves the system behaves as approved.
Is AI governance a legal requirement?
Partly. No law uses the term as a single duty, but the EU AI Act, GDPR and a growing number of national laws require pieces of it: AI literacy, human oversight, logs, risk assessment and transparency. Which pieces apply depends on your systems and your role.
What is the difference between AI governance and responsible AI?
Responsible AI describes the values: fair, transparent, accountable. AI governance is the machinery that makes those values happen: owners, approvals, rules and records.
Who should own AI governance in an organisation?
A senior sponsor owns the programme, and each AI system has its own accountable owner. Compliance, risk and the data protection officer support them. It should not sit with the IT team alone.
Does AI governance apply to AI tools we buy, not just ones we build?
Yes. If you deploy a bought system, you still carry the deployer’s duties. That includes AI features switched on inside software you already use.
Your next step
Today, open a spreadsheet and list every AI system your organisation uses, with one column for the owner. If any row stays blank, you’ve found where to start. To see AI agents, automations and approval gates working together, book a free walkthrough. A real person replies within one working day.
This article is general information, not legal advice.
Sources
- Regulation (EU) 2024/1689, the EU AI Act, EUR-Lex, published 12 July 2024. Articles 4, 5, 26, 50 and 99.
- Regulation (EU) 2016/679, the GDPR, EUR-Lex, 27 April 2016.
- ISO/IEC 42001:2023, AI management systems, ISO, December 2023.
- NIST AI Risk Management Framework 1.0, NIST, January 2023.
- OECD AI Principles, OECD, adopted 2019, updated May 2024.
- OutSystems State of AI Development research, 2026, via iTWire.
- Gartner predicts over 40% of agentic AI projects will be cancelled by end of 2027, Gartner, 25 June 2025.
- Regulation (EU) 2026/1744, the Digital Omnibus on AI, EUR-Lex, 2026.
- Texas Responsible Artificial Intelligence Governance Act (HB 149), Texas Legislature, signed 22 June 2025, in effect from 1 January 2026.
- South Korea’s AI Basic Act, IAPP. In effect from 22 January 2026.