Confident output can still be false

Generative AI may invent sources or merge reports about different people. An error in a hiring process can have significant consequences. Material findings must be checked against reliable, independent original evidence.

Never assume a matching identity

A shared name or photograph is insufficient to attribute a record or article to a candidate. Review appropriate identifiers, context and dates. Unverified matches must not be reported as established facts.

The danger of unexplained risk scores

An integrity or trustworthiness score can create a false appearance of precision. Its data inputs, weighting, limitations and potential bias must be understood. Avoid speculative predictions about loyalty, personality or future behaviour.

Automated decisions require safeguards

GDPR Article 22 establishes protections for certain solely automated decisions producing legal or similarly significant effects, subject to specified exceptions and safeguards. A rubber-stamp human signature is not meaningful intervention.

Prohibited and high-risk AI uses

The EU AI Act prohibits certain practices, including workplace emotion inference subject to narrow exceptions. Other recruitment systems may be high-risk. Annex III requirements apply from December 2027, but GDPR duties and applicable prohibitions are relevant already.


In the next article (part 4 of 5), we explore: “Human risk management in Spain: proportionate by design”.

Discuss responsible screening

Validato helps organisations design proportionate, documented, role-specific background checks. Contact our team to discuss your requirements.