A house newsletter · weekly on Tuesdays · researched and written by the Kurzia AI
Tuesday 17 November 2026 · Issue 6 · 5 min read
Sandbox rules adopted EU-wide; Article 27 assessment template out for comment; EDPB picks AI recruitment screening for 2027
The week's one thing
AESIA (Spanish AI Supervisory Agency) · 2 days ago
Implementing Regulation (EU) 2026/2487
Takeaway: 10 November 2026
The Commission has adopted an implementing regulation, dated 10 November 2026, that fixes common rules for the AI Act's regulatory sandboxes — the supervised environments where providers and would-be providers can develop, train and test systems before putting them on the market. Spain's AI agency AESIA announced it on 12 November. It governs every member state's scheme, whether national, regional or run jointly across borders.
What changes is harmonisation: shared access conditions and common project-selection criteria replace a patchwork of national arrangements, and SMEs, startups and small mid-caps get priority access plus regulatory support. Each project needs a defined test plan; supervisors may suspend testing where risks to health, safety or fundamental rights cannot be adequately mitigated; and taking part buys neither conformity nor a right to sell the system. Spain's pilot sandbox, run by SEDIA with the Commission, guided twelve high-risk systems across six sectors and produced sixteen practical compliance guides.
Deadlines and obligations
Sandbox implementing act adopted. Implementing Regulation (EU) 2026/2487, adopted 10 November 2026, sets EU-wide rules for the AI regulatory sandboxes member states must run for providers and prospective providers. AESIA (Spanish AI Supervisory Agency)
Omnibus prohibitions bite. From 2 December 2026 synthetic intimate imagery created without consent and AI-generated child sexual abuse material join the banned practices for providers and deployers. artificialintelligenceact.eu
FRIA template testing call. Annex III deployers have until 4 December 2026 to tell the AI Office they want to test the draft fundamental rights impact assessment template. European Commission
FRIA feedback window. The Commission's consultation on the draft Article 27 assessment template and its explanatory notice runs until 15 January 2027. European Commission
Article 27 duty. Annex III deployers must complete a fundamental rights impact assessment before deploying a high-risk system from 2 December 2027. European Commission
Guidance, codes and standards
Draft FRIA template — The Commission is collecting feedback on its draft template and explanatory notice for the Article 27 fundamental rights impact assessment that Annex III deployers must complete before putting a high-risk system to work. The AI Office drafted it with the Fundamental Rights Agency; once adopted, deployers must use it and notify the results to their market surveillance authority. Comments close 15 January 2027, and deployers can volunteer by 4 December 2026 to test it on a live system. This is the document your organisation will actually fill in for Article 27, and the consultation window is the chance to shape it before it becomes mandatory. European Commission
Prohibited-practices guidelines explainer — Worth carrying into a briefing: the text is not binding — the Court of Justice decides — but it says the prohibitions take direct effect in national courts, reach deployers as well as providers, and can catch general-purpose systems. It also reads the bans narrowly, given penalties of up to €35 million or 7% of worldwide turnover. The Guidelines are the clearest signal yet of how the Commission will treat the bans your organisation is already subject to, and of the two more arriving on 2 December 2026. The two Omnibus additions cover synthetic intimate imagery of real people made without their consent and AI-generated child sexual abuse material. artificialintelligenceact.eu (Future of Life Institute)
Enforcement and the authorities
EDPB November plenary — At its plenary on 11 November the EDPB adopted a statement on the first review of Korea's adequacy decision, finalised its processor-BCR recommendations — holders must align during application or in their 2028 annual update — and picked automated decision-making in recruitment, including AI-driven candidate screening, as its 2027 coordinated enforcement topic. The action will examine GDPR Article 22 and the right of access; national DPAs join voluntarily, launching during 2027 — the board's first concerted look at ADMT since 2018. If your organisation screens job candidates with AI, expect coordinated DPA scrutiny from 2027, on top of the AI Act's high-risk employment rules. A final report on this year's transparency action is due in early 2027. European Data Protection Board
Law 5321/2026 — Greece set its AI Act framework in Law 5321/2026, in force since its 20 July publication, and an IAPP analysis this week maps how it works: oversight spread across existing sectoral regulators rather than a new agency, with the Hellenic Data Protection Authority as market surveillance authority for prohibited practices, Annex III high-risk systems and Article 50 transparency, and as Greece's single contact point with the Commission. For anyone deploying AI systems in Greece, notifications and complaints under the AI Act and the GDPR now go to the same regulator. The bill was consulted from 21 June and enacted within two weeks of the consultation closing on 6 July. IAPP
