New rules on artificial intelligence and automated decision-making are making employers look harder at their recruitment tools. This article explains what the EU AI Act and the UK's updated data protection rules mean for psychometric testing, how traditional tests differ from AI scoring tools, and the practical questions employers should ask. It is a general guide, not legal advice.

Why the rules matter for recruitment

Recruitment decisions have a significant effect on people's lives, so regulators treat technology used in hiring with particular care. Two developments are especially relevant for employers in the UK and Ireland.

The EU AI Act. The EU's AI regulation (Regulation (EU) 2024/1689) classifies AI systems used for recruitment and selection, such as tools that screen, rank or evaluate candidates, as high-risk [1]. Draft guidance from the European Commission makes clear that tools can be high-risk where they materially influence employment decisions, even if a human makes the final decision. The high-risk obligations were originally due to apply from August 2026; proposals under discussion would move them for employment uses to December 2027, so employers should check the current position. The Act applies to employers using these systems in the EU, including UK organisations recruiting there.

UK data protection. The Data (Use and Access) Act 2025 updated the UK GDPR, including the rules on automated decision-making [2]. The changes are generally more flexible for employers, but come with safeguards: candidates must be told about significant automated decisions, be able to challenge them and ask for human involvement. In 2026 the Information Commissioner's Office published a report on automated decision-making in recruitment, and found that many employers wrongly assumed the rules did not apply because AI was "only supporting" human decisions [3].

Is a psychometric test "AI"?

Not necessarily. There is an important difference between two kinds of tool:

• Traditional psychometric tests score answers against a fixed key or scoring rules set by psychologists, and compare scores with a defined norm group. The same answers always give the same score, and the method can be explained and documented in a technical manual.

• AI scoring tools use models that infer results from data, such as scoring video interviews, CVs or free text with machine learning. Their logic can be hard to explain, and their outputs may change as models are updated.

Whether a particular tool falls within the AI Act depends on how it works, so employers should ask each supplier directly. Even where a tool is not "AI", equality law, data protection law and good practice still apply to how it is used.

Meaningful human involvement

Both sets of rules put weight on genuine human oversight. That means a person who understands the result, has the authority to reach a different decision and actually considers the candidate as an individual, not someone who simply confirms what the system recommends. Using test results as one source of evidence alongside interviews and other information, rather than as an automatic cut-off, is the simplest way to make that real.

Questions to ask about any assessment tool

• How are scores produced: by fixed rules or by an AI model?

• What does the test measure, and what evidence shows it is reliable, valid and fair for groups of candidates?

• Is there a technical manual or design document explaining how it was built and scored?

• Are any decisions made automatically, without a person reviewing the result?

• What do candidates receive: information about how their data is used, and feedback on their results?

• How are reasonable adjustments handled?

• Where is data stored, for how long, and is it anonymised afterwards?

How Viewpoint approaches this

Viewpoint's tests are scored with fixed, documented keys and rules, and results are compared with defined norm groups or external standards. Reports explain what each result means, and every candidate can receive a Candidate Report. Results are shown in bands from Very Low to Very High, with no pass marks: employers are encouraged to use them alongside interviews and other evidence, so people make the decisions. Candidate records are anonymised automatically after 18 months: names, contact details and other identifiers are removed, and only anonymous scores and demographic information are kept, to maintain our norm groups.

See our article on SJTs and fairness, our guide to AI and cheating in online tests, or contact us for our technical documentation.

Frequently asked questions

Does the EU AI Act apply to UK employers?

It can, where AI systems are used for recruitment in the EU, for example when hiring for roles in Ireland. Check with your legal adviser.

Are Viewpoint tests AI systems?

Our tests are scored by fixed, documented rules rather than machine-learning models, and people make the hiring decisions. Ask us for documentation if you need it for your own assessment.

Do we still need a human to review results?

Yes. Good practice, and in many cases the law, calls for meaningful human involvement in recruitment decisions.

Is this legal advice?

No. This is a general guide. Take legal advice on your own circumstances.

References

1. Regulation (EU) 2024/1689 of the European Parliament and of the Council laying down harmonised rules on artificial intelligence (Artificial Intelligence Act). Official Journal of the European Union.
2. Data (Use and Access) Act 2025, c. 18. London: The Stationery Office.
3. Information Commissioner's Office (2026). Report on the use of automated decision-making in recruitment. Wilmslow: ICO.