The European Commission adopted Commission Implementing Regulation (EU) 2026/2487 on 10 November 2026 and it was published in the Official Journal on 11 November 2026. The measure sets common rules for the establishment, operation and supervision of AI regulatory sandboxes under the EU AI Act.
Under the new implementing regulation, national competent authorities and the EU-level AI Office must apply uniform selection, participation and exit procedures for sandboxes; the regulation formally enters into force on 1 December 2026. The Commission says the purpose is to reduce fragmentation and give innovators a predictable, supervised environment to develop high‑risk systems. "AI regulatory sandboxes aim to enhance legal certainty for innovators and support them in achieving regulatory compliance," the Official Journal text states.
How the rules work
The implementing regulation lays out application and selection criteria, a mandatory "sandbox plan" describing test scope and safeguards, rules on fees and a written exit report market surveillance authorities must recognise. Participation may be free for SMEs and start‑ups where national authorities choose to waive charges; authorities must publish fee lists when calls open. The text makes clear that "leaving a sandbox is no approval to place the system on the market and grants no presumption of conformity."
Crucially for developers of automotive AI, the regulation gives supervised testers a conditional reduction in enforcement risk: a properly run sandbox project produces an exit report that member‑state market surveillance authorities and notified bodies must take into account when assessing conformity and may speed conformity procedures — but it does not replace conformity assessment or remove liability. The rules also allow participation where personal data are necessary for testing transport systems, subject to strict isolation and data‑protection safeguards.
What this means for the market
For Central and Eastern Europe the regulation matters for three reasons. First, the common rules make it easier for software teams, Tier‑1 suppliers and scale‑ups in CEE to choose where to test advanced driver‑assistance systems (ADAS), fleet‑automation software and other high‑risk applications without fearing a patchwork of national procedures. Second, the obligation for authorities to recognise exit documentation across the Union should speed up cross‑border conformity assessments and help CEE suppliers move from trial to volume production faster. Third, conditional protections and SME fee waivers lower the cost and regulatory risk of real‑world testing — a material advantage for smaller R&D centres in Poland, Slovakia, Hungary and Romania.
Implementation tasks fall to national competent authorities: each Member State must set up at least one recognised sandbox under the AI Act architecture and cooperating networks are expected to align before the Member‑State deadlines set in the Digital Omnibus. Companies planning to test systems that process real driving or personal data should plan for supervised sandbox applications and make sandbox plans detailed enough to preserve the implementing regulation's conditional shield.
Regulatory clarity will not remove the technical and capital hurdles to production‑grade autonomy, but the implementing regulation creates a faster, safer corridor from lab prototypes to conformity assessment for automotive AI — and CEE suppliers and auto‑software hubs should factor it into 2027 R&D planning.
Available vehicles on WorldTrader24 now →
Sources: Publications Office of the EU (Official Journal L 2026/2487), Quentir Systems analysis, Normdiff summary
