Spain's Audiencia Nacional, the national court that hears collective labour disputes, sanctioned an employer on 4 July for concealing how an algorithm that set shifts, days off and rest periods actually worked. Ruling 101/2026 concerns disclosure rather than dismissal, and legal analysis of it landed on 26 July1. It establishes that workers may know the logic of the system organising their week.
Spanish labour lawyers now expect the first rulings on dismissals justified by AI performance scoring within weeks. The remedy they anticipate is the ordinary one for unfair dismissal: reinstatement, back pay and compensation. An employer stays liable even where an outside vendor built the tool, which closes off the obvious defence of pointing at the supplier.
Brussels arrived at the same question from the regulatory side. The European Commission's draft guidelines interpreting Annex III of the AI Act, the list of uses the Act treats as high-risk, covering recruitment, pay, evaluation, monitoring and termination, closed for stakeholder feedback on Thursday 23 July2. They say a system counts as high-risk where its output "heavily influences" who advances or how a worker is evaluated, even when a human formally signs the decision. Few firms dismiss anyone by fully automated decision. Plenty score performance algorithmically and have a manager countersign. Brussels had already deferred these rules once.
Final guidelines are due at the end of 2026, ahead of a compliance deadline of 2 December 2027 that the Digital Omnibus pushed back from August this year. Germany's Bundesrat, the upper house of its parliament, approved the national enforcement law on 10 July, more than a year ahead of that deadline. A Madrid courtroom may reach a binding answer on AI dismissals before either date arrives.
