Examples
Worked examples
- Is an instance
A developer running a bias audit against a defined protected-attribute set to operationalise the HLEG "fairness" requirement into a checkable test.
Counter-examples
Looks similar, but isn't
- Not an instance
Citing a current US executive-order FLOPs reporting threshold — EO 14110 set one, but it was revoked in January 2025 and not replaced; only the EU AI Act Article 51 threshold is currently active.
Editorial commentary
The EU HLEG (2019) framework lists seven requirements for trustworthy AI: human agency and oversight; technical robustness and safety; privacy and data governance; transparency; diversity, non-discrimination, and fairness; societal and environmental well-being; accountability. The framework strongly influenced the EU AI Act. A parallel, intergovernmental framework — the OECD AI Principles, also adopted in 2019 — covers similar ground (human rights and democratic values, transparency, robustness and safety, accountability) at OECD/G20 scale rather than as an EU-specific instrument.
The current US/EU regulatory picture (a live area — check dates)
US Executive Order 14110 (October 2023) set a reporting threshold for dual-use foundation models trained above 1026 FLOPs. It was revoked on 20 January 2025 (by EO 14148) and replaced on 23 January 2025 by EO 14179, “Removing Barriers to American Leadership in Artificial Intelligence” — which carries no equivalent FLOPs-based reporting threshold. As a result, the EU AI Act’s Article 51(2) presumption — a rebuttable systemic-risk presumption triggered when a model’s cumulative training compute exceeds 1025 FLOPs (with Article 55 layering on evaluation, incident-reporting, and cybersecurity duties for models over that line) — is, as of this writing, the only FLOPs-based regulatory threshold still active among major jurisdictions. Any source still describing a current US executive-order FLOPs threshold is describing EO 14110, which no longer applies.
How organisations operationalise “trustworthy”
The HLEG’s seven requirements are a governance framework, not a testing method — organisations translate them into practice through the NIST AI Risk Management Framework (Govern/Map/Measure/Manage functions), documented AI assurance processes, an AI safety case for higher-risk deployments, and targeted checks such as a bias audit or a broader model audit. A HLEG requirement like “fairness” is not itself measurable; a bias audit against a defined protected-attribute set is what makes it checkable.
Related pages
See also NIST AI RMF, AI assurance, AI safety case, model audit, bias audit, compute FLOPs estimation, and frontier model.
References
- EU High-Level Expert Group on AI, “Ethics Guidelines for Trustworthy AI” (2019); EU AI Act (Regulation 2024/1689), Articles 51, 55.
- US Executive Order 14110 (2023, revoked 2025); Executive Order 14179 (23 January 2025).
- NIST AI Risk Management Framework (AI RMF 1.0), nist.gov.
Also known as
EU Trustworthy AI · HLEG Trustworthy AI
Machine-readable encodings
Use in your systems
<role vocab="credit"
vocab-identifier="https://casrai.org/dictionary/"
vocab-term="Trustworthy AI"
vocab-term-identifier="https://casrai.org/dictionary/term/trustworthy-ai" />{
"@context": "https://schema.org",
"@type": "DefinedTerm",
"@id": "https://casrai.org/dictionary/term/trustworthy-ai",
"name": "Trustworthy AI",
"identifier": "https://casrai.org/dictionary/term/trustworthy-ai",
"description": "AI systems exhibiting properties (lawful, ethical, technically robust) that warrant the trust of users, affected parties, and society, as articulated in the EU High-Level Expert Group's framework and adopted in subsequent regulation.",
"inDefinedTermSet": "https://casrai.org/dictionary/domain/ai-ml-research-outputs#set",
"url": "https://casrai.org/dictionary/term/trustworthy-ai",
"sameAs": [
"EU Trustworthy AI",
"HLEG Trustworthy AI"
],
"license": "https://creativecommons.org/licenses/by/4.0/",
"publisher": {
"@id": "https://casrai.org/#organization"
},
"author": {
"@id": "https://casrai.org/#editorial-team"
},
"datePublished": "2026-05-21T02:22:50",
"dateModified": "2026-08-23T05:20:13",
"inLanguage": "en-GB",
"isAccessibleForFree": true
}
