SLOVENIA Trends and Developments Contributed by: Vesna Cukrov and Veronika Cukrov, Law Firm Cukrov (Odvetniška družba Cukrov o.p., d.o.o.)
The EU Artificial Intelligence Act (Regulation (EU) 2024/1689) classifies as high-risk, at point 8 (a) of Annex III, AI systems intended to be used by a judicial authority or on its behalf to assist in researching and interpreting facts and the law and in applying the law to a concrete set of facts – adding expressly, “or to be used in a similar way in alternative dispute resolution”. Recital 61 elaborates that AI systems intended to be used by alternative dispute resolution bodies for those purposes should be considered high-risk where the outcomes of the alternative dispute resolution pro - duce legal effects for the parties. It continues that AI tools can support the decision-making power of judg - es, but must not replace it, and final decision-making must remain a human-driven activity. Whether a private commercial arbitrator seated in Ljubljana falls within that category is unresolved. An arbitrator is not a judicial authority; but the phrase “in a similar way in alternative dispute resolution” is broad, and an arbitral award unquestionably produces legal effects for the parties. An arbitrator deploying a legal-analysis tool could thus be a deployer of a high- risk system, with the attendant obligations of human oversight, transparency and record-keeping. The timetable has shifted. The Digital Omnibus on AI, adopted by the European Parliament on 16 June 2026 and by the Council on 29 June 2026, deferred the application of the high-risk obligations for standalone Annex III systems from 2 August 2026 to 2 December 2027, and for AI embedded in regulated products to 2 August 2028. That is a material reprieve, but not from everything. Most transparency obligations under Article 50 con - tinue to apply from 2 August 2026, while the marking requirement for systems already on the market has moved to 2 December 2026. The European Commis - sion published guidance on the implementation of these transparency obligations on 20 July 2026. Slovenia has moved faster than many member states on national implementation. The Act implementing Regulation (EU) 2024/1689 (ZIUDHPUI) was adopted on 23 October 2025, published in Official Gazette No 85/25, and has applied since 21 November 2025. It
designates the competent authorities, distributes market surveillance functions among several bodies, establishes a five-member National Council for Eth - ics in Artificial Intelligence and provides for regulatory sandboxes, the first of which is to be established by 2 August 2026. Slovenian businesses therefore already have a supervisory architecture in place, well ahead of the substantive high-risk obligations taking effect. Evidence: fabrication, deepfakes and the problem of proof ARIHQ concerned fabricated law. The more challeng - ing issue now emerging relates to fabricated fact. Courts have already begun sanctioning counsel for unverified AI use. In Ayinde v The London Borough of Haringey (2025) EWHC 1383 (Admin), the English High Court emphasised that the use of AI requires appropriate oversight. In the Abu Dhabi Global Mar - ket, the court in Arabyads v Gulrez Alam (2025) ADG - MCFI 0032 ordered wasted costs against a legal team that filed false AI-generated authorities, characteris - ing the conduct as reckless. The Tribunal adminis- tratif d ’ Orléans admonished counsel on 29 December 2025 (No 2506461) over numerous false references. The Qatar International Court and Dispute Resolu - tion Centre issued Practice Direction No 1 of 2026, requiring lawyers to verify AI-generated submissions independently. The next frontier is documentary evidence. Genera - tive tools now produce convincing audio, images and video, and fabricate entirely ordinary-looking docu - ments – an internal email, a signed delivery note, a site photograph. Arbitration is more exposed here than court litigation because a party producing a forged document rarely faces criminal sanction, and because tribunals seldom possess the machinery to test authenticity properly. Parties can reduce that exposure by: • preserving native files with intact metadata rather than working from printouts or converted copies; • raising authenticity concerns early, in the first pro - cedural order, rather than at the hearing once the timetable has hardened;
876 CHAMBERS.COM
Powered by FlippingBook