Transparency in AI

In short

Transparency in AI means being open about when AI is being used, what it can and cannot do, how it was built and how it reached a result, so people can make informed choices and challenge decisions that affect them.

Also called: AI transparency, Algorithmic transparency
Topic: Safety, ethics and risk
Level: Beginner
First described: Transparency and explainability is Principle 1.3 of the OECD AI Principles, adopted in May 2019, the first intergovernmental standard on AI⁶¹.

Transparency covers several different things. The OECD AI Principles, first adopted in 2019 and updated in May 2024, ask organisations to help people understand AI systems' capabilities and limits, make people aware when they are interacting with AI, explain the main factors behind a result where feasible, and give those affected a way to challenge it¹⁶.

It does not usually mean publishing source code or training data. The OECD says transparency "does not in general extend" to disclosing source code or proprietary datasets, which may be too technical to be useful and may be protected as trade secrets¹.

In practice, transparency shows up as model cards and system cards published by AI companies, labels on AI-generated images, and notices telling you a chatbot is not a human². Explainability is a related but narrower idea: giving plain information on the factors and logic behind a particular result¹.

Some of this is becoming law. The EU AI Act requires chatbots to tell people they are talking to an AI unless it is obvious, requires AI-generated content to be marked in a machine-readable way, and requires anyone publishing a deepfake to disclose it, with lighter rules for clearly artistic or satirical work². These duties apply from 2 August 2026².

In the UK, central government departments must publish records under the Algorithmic Transparency Recording Standard (ATRS), explaining how and why they use algorithmic tools³⁴. The government's public register held 152 records in October 2026⁵.

Example

The Ministry of Justice published an ATRS record in September 2026 for PolyAI, an AI voice assistant that handles routine phone calls, so the public can see what the tool does and why it is used⁵.

Why it matters

You cannot question a decision you do not know a machine made. Transparency is the first step to holding AI users, including governments, to account.

Questions people ask

Is transparency the same as explainability?

Not quite. Transparency is the broad duty to be open about AI use, capabilities and limits. Explainability is one part of it: giving plain information on the factors and logic behind a specific result¹.

Do UK public bodies have to say when they use AI?

Central government departments must publish Algorithmic Transparency Recording Standard records for in-scope tools, and the government says it intends to extend this more widely across the public sector⁴. Records are published on GOV.UK⁵.

Do chatbots have to say they are AI?

Under the EU AI Act, yes, from 2 August 2026, unless it is obvious from the context. Providers must design chatbots so people are informed they are interacting with an AI system².

Related terms

Sources

Back to the glossary