AI ethics: who bears responsibility?
Accuracy is not the only value; data carries history; proportionate transparency; responsibility stays with people; governance as a cycle.
Accuracy is not the only value
A system that scores well on average may fail for a minority or in rare high-stakes cases. Ethics weighs accuracy against privacy, fairness, safety and the right to contest outcomes.
Data carries history
Datasets reflect what was measured and who was included. Past discrimination can be learned and amplified. Removing a sensitive label is not always enough if proxies remain.
Proportionate transparency
Not every user needs model equations; they need to know a system was used, what role it played, and how to appeal. Transparency is context-sensitive explanation, not dumping secrets or personal data.
Responsibility is not assigned to the machine
People and institutions choose goals, data, deployment and oversight. When harm occurs, responsibility tracks those choices — not a slogan that “the AI decided”.
Governance is an ongoing cycle
Impact assessment, monitoring, incident response and updates matter as much as a one-time launch checklist.
A practical example
A hiring screen that ranks CVs must be auditable for disparate impact and contestable by candidates — accuracy alone is an incomplete brief.
For learners
Ask of any tool: who is affected, who can override, and what evidence would trigger pausing use.
Sources
Pair this page with independent institutional guidance as linked when available on the Arabic original and updates.
In practice
This English page adapts the Arabic forum article on “AI ethics: who bears responsibility?” without adding unsupported claims. In practice, treat this explainer as a working note rather than a finished verdict. Re-read the key section once, open at least one independent source from the list when available, and write three lines in your own words: what the claim is, what evidence supports it, and what would change your mind. If you teach or share the idea, keep the distinction between fact and interpretation visible — the same standard we use on Vemetto’s Arabic originals. Returning to the practical example a week later often reveals what you actually retained.
Sources
- UNESCO — Recommendation on the Ethics of AI ↗
- OECD AI Principles ↗
- NIST AI Risk Management Framework ↗
Original Vemetto explainer in English, adapted from our Arabic page. Last reviewed: 9 November 2026.
Questions and practical notes
After reading
- Note one idea in your own words
- Open one source link
- Apply the practical example if present
Frequently asked questions
Is this a machine translation?
No. It is an English adaptation of the Arabic explainer, keeping the same claims and sources.
Where are the sources?
Listed on this page when the Arabic original cites open URLs; always prefer primary links.
How long should I spend?
Read once, then try the practical example or discussion prompt — that is the point of the forum style.