For Heads-of · Practitioner
Transparency
Also known as: AI transparency
The disclosure of information about an AI system to the people who need it: that AI is being used, what it does, what it is for and what risks it poses, in terms they can understand. For people whose personal data it uses, being informed is also a legal right.
- transparency
- disclosure
- consent
- Article 50
- right to be informed
Definition
Transparency is about openness: users, regulators and affected people have access to factual information about the AI systems that affect them, in plain language rather than technical detail. A system can be transparent without being explainable; explainability is an account of why it produced a specific output, and an organisation usually needs both. The absence of the second is inadequate explainability for affected individuals.
Disclosure duties
The EU AI Act requires people to be told when they are interacting with an AI system or seeing AI-generated content (Article 50), and requires providers of high-risk systems to give deployers instructions for use (Article 13). Under UK GDPR, people have a right to be told how their personal data is used, including whether decisions about them are automated. NIST's Generative AI Profile calls for disclosure of capabilities and limitations. AI transparency and disclosure labelling is the control.
Transparency and consent
Consent is one lawful basis for processing personal data, not the only one, and for many AI uses another basis is more appropriate. Where consent is relied on, UK GDPR requires it to be freely given, specific, informed and unambiguous, and as easy to withdraw as to give. Transparency is required whatever the lawful basis: telling people is not the same as asking them. The register entry for failing to tell people they are dealing with AI is undisclosed AI interaction.
Related Terms
Accountability
The principle that individuals and organisations can be held responsible for decisions made by the AI systems they deploy, and must be able to explain those decisions to the people affected.
Explainability
The ability to give a person an account of why a system produced a particular output, in terms they can act on. Explainability is a property of the account and its audience, not of the model: an opaque model can be explainable, and an interpretable one can fail to be explained.
Model Card
A document describing a machine learning model's intended uses, performance, limitations and known risks, so that people deciding whether and how to use it can do so on evidence. System cards extend the idea to the deployed system and its safeguards, and EU AI Act documentation duties now require much of the same content.
Related risks and controls
Inadequate explainability for affected individuals
Transparency's harder obligation: explaining a decision to the person it affected.
Undisclosed AI interaction
The failure mode when consent and transparency are not built in from the start.
AI transparency and disclosure labelling
The control that makes AI use visible by default.