AI governance in UK employment & hr.

Automated hiring rejections, AI performance scores triggering discipline, and algorithmic redundancy selection engage the Articles 22A-22D safeguards that replaced Article 22 on 5 February 2026: individuals must be given information about the decision, an opportunity to make representations, access to human intervention, and a route to contest it. Human review must be genuine, rubber-stamping AI recommendations does not satisfy the requirement.

AI tools producing outcomes that disproportionately disadvantage protected groups may constitute indirect discrimination regardless of intent. Bias testing before deployment is now an expected baseline.

Employee monitoring AI requires lawful basis, DPIA for systematic monitoring, and transparent advance notice. Productivity monitoring without specific disclosure is unlawful under UK GDPR.

Acas issued informal advice for employers on using AI at work in 2025, covering clear AI policies, early consultation with staff and unions, checking AI outputs for accuracy, and applying data protection policies. It is advisory good practice, not an Acas statutory Code of Practice, so it carries no automatic evidential weight in Employment Tribunal proceedings. for consultation, transparency, and human oversight that Employment Tribunals will reference in unfair dismissal and discrimination cases.

Employers must consult before introducing AI that significantly changes working conditions. AI-assisted redundancy selection requires objective, transparent criteria that employees can understand and challenge.

Organisations with 50+ employees have information and consultation rights before significant AI-driven changes to employment conditions under the Information and Consultation of Employees Regulations 2004.