Right to Explanation

Official/legal formulation. The EU Artificial Intelligence Act expressly creates a targeted right to explanation for affected persons: where a deployer’s decision is taken based mainly on the output of a high‑risk AI system and produces legal effects or similarly significantly affects the person, the affected person "shall have the right to obtain from the deployer clear and meaningful explanations of the role of the AI system in the decision‑making procedure and the main elements of the decision taken." (EU AI Act, Article 86). ([eur-lex.europa.eu](https://eur-lex.europa.eu/eli/reg/2024/1689/oj))

What the term covers (context and scope). In regulatory and standards language the right to explanation sits at the intersection of transparency, explainability, and remedies: it is a procedural right for individuals to receive information that helps them understand how and why an AI‑informed decision affected them, and to enable challenge or redress. The EU formulation is risk‑targeted (applies to high‑risk AI outputs that produce legal or similarly significant effects) and is subject to existing Union or national law exceptions. Parallel policy and standards efforts (NIST, OECD, UNESCO, ISO) treat explainability and transparency as characteristics of trustworthy AI systems that support such rights in practice by requiring information, documentation and user‑appropriate explanations. ([eur-lex.europa.eu](https://eur-lex.europa.eu/eli/reg/2024/1689/oj))

Jurisdictional variations.

• European Union (primary legal source): The AI Act gives a concrete, enforceable right against deployers in the high‑risk context (Article 86), framed to be "clear and meaningful" and limited where other Union law provides the same right or where exceptions apply; the Act’s recitals clarify the right’s purpose is to let individuals exercise remedies and understand impacts. This is a distinct, targeted right rather than a universal catch‑all entitlement. ([eur-lex.europa.eu](https://eur-lex.europa.eu/eli/reg/2024/1689/oj))
• United States (policy & guidance): U.S. federal policy and guidance (e.g., the White House Executive Order on safe, secure and trustworthy AI) emphasise transparency, testing, and oversight but do not create an identically worded statutory "right to explanation"; agencies (NIST, FTC) promote explainability/transparency as characteristics of trustworthy AI and consumer protection obligations (FTC guidance) that require explainable, non‑deceptive uses of AI and may force businesses to disclose or explain automated outcomes under existing consumer or sectoral laws. Practically, U.S. enforcement tends to rely on existing statutes (consumer protection, fair credit, anti‑discrimination) and guidance rather than a freestanding private right to explanation. ([bidenwhitehouse.archives.gov](https://bidenwhitehouse.archives.gov/briefing-room/presidential-actions/2023/10/30/executive-order-on-the-safe-secure-and-trustworthy-development-and-use-of-artificial-intelligence/))
• International / standards bodies: OECD principles and UNESCO recommend transparency and explainability as duties of AI actors, encouraging meaningful information to enable challenge; ISO/IEC 22989 provides standardized terminology: "explainability" as a property enabling humans to understand important factors influencing outputs. These instruments are influential soft‑law and standards references that shape how jurisdictions operationalize the right to explanation (e.g., what counts as a meaningful explanation, audience tailoring, traceability). ([oecd.org](https://www.oecd.org/en/topics/ai-principles.html?utm_source=openai))

Practical implications for businesses operating across jurisdictions. Companies must translate the legal or policy requirement into operational processes: risk assessment to identify when a decision triggers the right; documentation and recordkeeping to produce the required explanation; role clarity (provider v. deployer) to determine which actor must respond; tailoring of explanations to the recipient; and legal checks for trade secrets, privacy, or national security limits. In the EU a deployer must respond to a request under Article 86 for high‑risk AI outputs; in the U.S. firms may face FTC enforcement or sectoral notice/appeal obligations (e.g., FCRA adverse action notices) requiring disclosure of factors used in automated decisions. Standards (ISO) and guidance (NIST, OECD, UNESCO) inform what explanations look like in practice (feature importance, counterfactuals, human‑readable rationale, process descriptions). ([eur-lex.europa.eu](https://eur-lex.europa.eu/eli/reg/2024/1689/oj))

Key requirements and good‑practice criteria. Implementing a defensible right‑to‑explanation practice typically involves the following elements (technical, organisational and legal):

• Identification: classify decisions and AI systems to determine whether a request is in scope (e.g., high‑risk list in EU AI Act).
• Documentation & traceability: maintain logs, datasets, model versions, pre‑deployment testing and impact assessments to support explanations and audits.
• Audience‑appropriate explanations: provide clear, meaningful explanations suitable to the requestor (plain language summaries, key factors, counterfactuals, and the AI system’s role), following NIST/ISO‑aligned guidance about explainability and interpretability. ([airc.nist.gov](https://airc.nist.gov/airmf-resources/airmf/3-sec-characteristics/?utm_source=openai))
• Legal balancing: screen for limits (privacy, trade secrets, security) and for overlapping legal rights; apply lawful exceptions where authorized but document the basis for non‑disclosure.
• Governance and redress: routes for appeal, human review, remediation and updates to models/processes where explanations reveal error or discrimination.

Examples. A bank deploys a high‑risk credit decision model in the EU: if a loan is denied and the decision was based mainly on a high‑risk AI output, the applicant can request an explanation from the deployer explaining the AI’s role and the main elements of the decision (EU AI Act, Article 86). In the U.S., the same denial would trigger FCRA/ECOA notice obligations and FTC scrutiny; firms should disclose key factors and provide an adverse‑action notice where applicable and be prepared to show testing and audits to regulators. Standards‑level implementations often use counterfactual explanations ("If X had been Y, the outcome would have changed") and feature‑importance summaries to meet the practical goal of enabling contestation. ([eur-lex.europa.eu](https://eur-lex.europa.eu/eli/reg/2024/1689/oj))

Cross‑references. See related concepts: explainability and interpretability (NIST and ISO definitions), transparency and traceability (OECD, UNESCO, ISO), automated decision‑making / ADM and data protection rights (GDPR Art.13–15, Art.22), and human oversight / meaningful human intervention (EU AI Act human‑oversight rules). Implementations of the right to explanation should be coordinated with privacy law, trade‑secret protection and sectoral notice/appeal regimes. ([airc.nist.gov](https://airc.nist.gov/airmf-resources/airmf/3-sec-characteristics/?utm_source=openai))