Glossary
Trustworthy AI
A policy term, especially in European documents, for AI that is lawful, ethical and technically robust.
Last reviewed 2026-08-22
In plain language
The European Commission’s High-Level Expert Group popularized “trustworthy AI” through ethics guidelines that list requirements such as human agency, privacy, transparency, diversity, societal well-being and accountability. Trustworthiness is a goal, not a certification by itself. Later EU law uses related but distinct legal categories (for example, high-risk AI systems) rather than treating “trustworthy” as a statutory status. The relationship between the two is a common point of confusion: high-risk AI is a defined legal category under the EU AI Act, triggering specific binding obligations for systems used in areas like employment, credit or law enforcement, while trustworthy AI is an ethics-guideline vocabulary with no independent legal force of its own—a system can be legally classified high-risk and subject to strict obligations without ever being described as “trustworthy,” and a vendor can call a low-risk product “trustworthy AI” as marketing language with no external check on the claim. The term is also often used interchangeably with “responsible AI,” but the two have different origins and audiences: “trustworthy AI” traces specifically to the High-Level Expert Group's requirements list, while “responsible AI” is a broader, less formally defined label used more often by companies and in US policy discussion, without a single reference document behind it. Readers encountering “trustworthy AI” in an EU policy document should expect it to point back to that specific requirements list, not to function as a general synonym for “safe” or “compliant.”
Why it matters for AI governance
The term appears in both ethics guidance and political communication. It should not be read as a guarantee that a system is safe, lawful or rights-respecting.
Authoritative sources
- Ethics Guidelines for Trustworthy AI (opens in a new tab) — European Commission