The short answer
Ethics asks what we should do. Safety asks how a system fails and how to stop it failing badly. Governance asks who decides, under what rules, with what accountability. A text can be strong on one and silent on the other two.
Responsible AI, Human Rights and Safety · Starter
- In one sentence
- AI ethics is the practice of checking whether AI systems are fair, safe, accountable and respectful of human rights before and during their use.
- Why it matters
- AI systems can shape who gets information, services, attention or opportunities. If the design or use is careless, people can be treated unfairly, exposed to privacy risks, or pushed into decisions that should stay with humans. AI ethics gives non-specialists a way to ask better questions before harm occurs: What is this system for? Who might be missed or misjudged? What evidence supports its use? Who is accountable? What human review is needed? These questions help organisations make decisions that are lawful, defensible and more trustworthy in practice.
- The everyday version
- AI ethics means checking whether an AI tool should be used, who it could help or harm, and what human safeguards are needed.
- The one thing to remember
- AI ethics asks whether an AI use is justified at all, then requires evidence, safeguards, accountability and continuing review, not a one-off statement of good intentions.
- The commonest misconception
- AI ethics is just a philosophical debate with no effect on real systems. Ethical concerns become practical choices: what data can be used, which tasks are off-limits, what testing is required, when a person must review outputs, and whether a system should be deployed at all. In organisations, these choices shape procurement, design, approval and monitoring.
Read the full page
Responsible AI, Human Rights and Safety · Starter
- In one sentence
- AI safety means designing, testing and using AI systems so they are less likely to cause harm, especially in situations where people could be affected.
- Why it matters
- Non-specialists often meet AI as a convenient tool, but convenience can hide risk. A system can sound confident while being wrong, miss important cases, produce harmful advice, or push people to trust it too much. In low-risk tasks, that may waste time. In higher-risk tasks, it can affect safety, rights, services, reputation or access to help.
AI safety helps organisations ask practical questions before relying on a system: What could go wrong? Who could be harmed? What checks are needed? When must a human make the final call? These questions support better decisions about whether to use AI, where to limit it, and when not to use it at all.
- The everyday version
- AI safety means making sure an AI tool is less likely to harm people and that humans stay in control where the stakes are high.
- The one thing to remember
- AI safety is the continuing work of finding possible harms, testing failure modes, setting limits, keeping appropriate human oversight and responding when problems emerge.
- The commonest misconception
- If an AI system works well most of the time, it is safe. Safety is about the consequences of failure, not only average performance. A system that is usually correct may still be unsafe if it fails in high-risk cases, fails unpredictably, or encourages people to rely on it without proper checks.
Read the full page
AI Governance and UN Terminology · Starter
- In one sentence
- The rules, institutions and processes used to guide how AI is built, used and overseen so its benefits and risks are handled in the public interest.
- Why it matters
- Non-specialists often meet AI governance before they meet advanced AI. It shapes what can be bought, deployed, restricted, reported and improved. If you help choose software, approve budgets, design services, draft policy or explain risks to leadership, governance determines the questions you should ask: Who is responsible? What checks exist? What evidence supports this system? How will problems be found and fixed? Without governance, AI use can become unclear, uneven and hard to challenge. With governance, decisions are more traceable, and organisations are better able to balance innovation, safety, rights and public trust.
- The everyday version
- AI governance is the set of rules and responsibilities around AI. It covers who may use it, what checks are required, and what happens if it causes problems.
- The one thing to remember
- AI governance is the surrounding system of rules, responsibilities and checks for building, buying and using AI, including what happens when the system fails.
- The commonest misconception
- AI governance just means passing an AI law. Laws are only one part of governance. Governance also includes standards, procurement rules, oversight bodies, audits, reporting processes, professional guidance and practical ways to correct problems. An organisation can do some governance even before a new law exists.
Read the full page
Responsible AI, Human Rights and Safety · Starter
- In one sentence
- An approach to designing, using and governing AI so it is safe, fair, accountable and under appropriate human control.
- Why it matters
- Responsible AI affects whether a tool saves time safely or creates unfair, unsafe or unaccountable outcomes. For non-specialists, it helps frame better questions before adoption: Is this use appropriate? Who might be harmed? What evidence supports it? Who checks the output? What happens when it fails? These questions improve procurement, policy, leadership decisions and day-to-day use.
- The everyday version
- Responsible AI means using AI with safeguards so people are treated fairly, risks are checked, and humans remain accountable for important outcomes.
- The one thing to remember
- Responsible AI means managing an AI system throughout its lifecycle, with documented safeguards, testing, accountability, monitoring and meaningful human control.
- The commonest misconception
- If an AI system is accurate, it is responsible. Accuracy matters, but it is only one part. A system can score well on average and still be unfair to some groups, expose private data, fail in unusual conditions or be used without clear accountability. Responsible AI asks about the whole system around the model, not just one performance number.
Read the full page