News · 26 June 2025
The Data (Use and Access) Act and automated decisions: what changes in the UK
The Data (Use and Access) Act 2025 received Royal Assent on 19 June 2025. It amends several areas of UK data law, including the UK GDPR rules on automated decision-making, which matter to anyone deploying models that affect individuals.
Where the law was
Under the previous UK GDPR approach, individuals had a general right not to be subject to decisions based solely on automated processing that produce legal or similarly significant effects, subject to limited exceptions. That wording was widely seen as broad and hard to apply.
What the new framework does
The Act replaces that structure with a set of provisions that allow significant decisions based solely on automated processing in more situations, while requiring safeguards. For decisions that use special category data, such as health data, the conditions remain stricter. Where automated decision-making is used, organisations are expected to provide:
- Information about the decision-making,
- the opportunity to make representations,
- access to human intervention, and
- a way to contest the decision.
The Act also clarifies what counts as a decision based solely on automated processing: it is one with no meaningful human involvement. Rubber-stamping a model's output is not meaningful involvement.
What this means for explainability
Safeguards of this kind need information to work. To contest a decision, a person needs to know the main factors behind it. To intervene meaningfully, a human reviewer needs to see the evidence behind the model's output and have the time and authority to disagree.
So even though the Act relaxes the starting point, it increases the importance of well-designed human-in-the-loop systems. A model that cannot explain itself makes meaningful review impractical, and "meaningful" is exactly the word the law relies on.
A checklist for teams
- List decisions your systems make or recommend about individuals.
- For each, decide whether a human genuinely reviews it, and document how.
- Prepare plain-language explanations of the main factors.
- Build a contest-and-review route, and measure how often reviewers overturn the model. A rate of zero should worry you.
- Update privacy notices, and check that data protection impact assessments cover these flows.
This is a high-level view, not legal advice. Review specifics with counsel, particularly for health or employment contexts.
Written in June 2025; details reflect the situation at that time.