AI at work, by profession
AI in recruitment: what UK law says about automated sifting, scoring and rejection
What changed on 5 February 2026, what the Information Commissioner's Office found when it looked at employers, and why removing sex and ethnicity from the data does not remove discrimination.
Checked against the published sources on 2 October 2026.
The short answer
Since 5 February 2026 UK law permits a solely automated decision with a significant effect, such as rejecting a job application, on ordinary personal data, provided safeguards are in place. The candidate must be told, and must be able to make representations, obtain human intervention and contest the decision. The Equality Act 2010 applies to the outcome whoever, or whatever, made it, and the employer is liable.
- Law changed
- 5 February 2026, under the Data (Use and Access) Act 2025
- ICO report
- 31 March 2026, on automated decisions in recruitment
- Main finding
- Employers underestimated when tools were deciding
- Liability
- The employer, whoever built the tool
When the tool is really deciding
Suppliers almost always describe their product as decision support. The ICO found that many such tools showed no meaningful human involvement in practice. Its draft guidance says involvement must be active rather than tokenistic: a trained reviewer with the information, the time and the authority to reach a different outcome. A person who built the system does not count.
A simple test: count how often a reviewer changes the tool's outcome. A figure of zero tells you who is deciding.
Discrimination without intent
A gap in employment can stand in for maternity or illness, a postcode for ethnicity, a graduation year for age. A tool trained on past hiring learns past preferences. A neutral looking criterion that puts a protected group at a particular disadvantage is indirect discrimination unless it can be justified.
The only way to know is to compare who applies with who passes each automated stage, and to have a person review a sample of those rejected.
Reasonable adjustments
A timed online test or a video interview scored on speech or eye contact can put disabled candidates at a substantial disadvantage. There has to be a clearly offered way to ask for an adjustment before the automated stage, and a human route for anyone who needs one.
This page explains what the rule says. It is not legal, regulatory, clinical or financial advice on your own position. Rules change, so read the sources listed beside it, and tell us at hello@wajd.co.uk if something here is out of date.
Learn it properly, free
This guide is the summary. The course teaches it in full, with a video conversation, worked scenarios and a knowledge check after each module. Every module is free to read with no account. A certificate, if you want one once you pass, is £9.
AI at work, by industry · 1.5 CPD hours
AI in HR and recruitment: fair, lawful and open to challenge
Questions people ask
Is it legal to reject candidates automatically in the UK?
It can be, with safeguards. The candidate must be given information about the decision and be able to make representations, obtain human intervention and contest it. Stricter conditions apply where special category data is used.
Who is liable if a recruitment tool discriminates?
The employer. The Equality Act applies to the outcome, and 'the software chose' is not a defence in an employment tribunal.
Does a manager approving the shortlist count as human review?
Not if they approve it without the information, time or authority to reach a different outcome. Involvement has to be active rather than tokenistic.
Does the EU AI Act affect UK recruiters?
It is not UK law. If you recruit in the European Union, it treats AI used for recruitment as high risk, with duties applying from 2 December 2027, and it already bans emotion recognition in the workplace except for medical or safety reasons.