MMOOZZGG.com

Glossary

What are AI ethics principles?

In short

AI ethics is a set of principles for building and using artificial intelligence so that it does not harm people: fairness, transparency, privacy and human responsibility.

Security and law · AI ethics

MMOOZZGG.com editorsChecked: 1 min read

For a child

A strong child can help younger kids, or push them. Strength itself is not good or bad: what matters is how you use it. Smart programs are strong too: they count fast, write and recognize faces. AI ethics is an agreement among people on how to use this strength honestly and carefully, without hurting anyone.

In plain words

A neural network does what it was trained and told to do, not what is right. So people agree on rules in advance. Do not decide a person’s fate in the dark: they must understand why they were refused. Do not sort people into convenient and inconvenient because of skewed data. Do not collect more about people than you need. Tell people when a program is talking to them. And most important: a person is responsible for the machine’s decisions. These rules are written into international documents, laws and internal company codes.

For business

For a company, ethics turns from philosophy into risk. An unfair refusal to a client, hidden data collection or a bot that pretends to be a person leads to complaints, inspections and a loss of trust. Federal Law 243-FZ, Russia’s AI law, names among its principles of regulation the protection of human rights and freedoms, respect for a person’s autonomy and free will, and a risk-based approach. In practice, this is a set of concrete rules: which decisions AI makes alone and which it only proposes; which data must not go to the model; how a client learns they are talking to a bot; who reviews disputed cases. Such rules are collected into an AI use policy and discussed with managers in training.

For a specialist

AI ethics studies the norms for building and using AI systems and the ways to check them. The UNESCO Recommendation on the Ethics of AI, adopted in November 2021 by all 193 member states, sets four values and ten principles, among them proportionality and do no harm, privacy, transparency and explainability, human oversight, and fairness and non-discrimination. The OECD AI Principles (2019, updated in May 2024) include five value-based principles, and 47 adherents have joined them. The engineering side: bias audits and fairness metrics, explainability, impact assessment, documentation of models and data, adversarial testing, and ways to appeal decisions. The difference from alignment: alignment solves the technical task of making a model match set goals, while ethics asks which goals and limits are acceptable and who answers for them. The principles are binding only where a law or a contract sets them down.

Example

A clinic chain introduces AI to sort patient requests. By its internal rules, the model only proposes the urgency, the doctor on duty makes the final decision, and the patient is told in the chat right away that a program gives the first answers.

Related terms

Questions and answers

Which principles make up AI ethics? The ones named most often are fairness, transparency and explainability, privacy protection, safety, and human responsibility for decisions. International recommendations, laws and company codes set them down.
How does AI ethics differ from an AI law? A law is binding, and breaking it is punished. Ethical principles are broader and softer: a company adopts them voluntarily, but some of them gradually turn into laws.

Sources

Where companies use it

Not sure whether your company needs this?

AI solutions map
Contact us