Commission Report on the review of prohibitions and high-risk AI (Article 112(1))

Tag: S-2026-05-22-eu-ai-office-prohibitions-highrisk-review Type: report (European Commission monitoring report) Author(s): European Commission (EU AI Office context) Date of source: 2026-05-22 (publication date; page last updated 26 May 2026) Date ingested: 2026-06-08 Authority weight: high — primary EU institution discharging a statutory review obligation under the AI Act. Raw file: S-2026-05-22-eu-ai-office-prohibitions-highrisk-review.md. External URL: https://digital-strategy.ec.europa.eu/en/library/report-review-prohibitions-and-high-risk-ai

What it claims

On 22 May 2026 the Commission adopted a report assessing whether the list of prohibited AI practices (Article 5) and high-risk use cases (Annex III) of the EU AI Act need amending [S-2026-05-22-eu-ai-office-prohibitions-highrisk-review]. The report is a monitoring tool for the annual review required under Article 112(1), produced using the “objective and participatory methodology” the Article requires, with Member State and stakeholder input via the AI Board and public consultation. Its central conclusion is that substantive evaluation is still early-stage: the Article 5 prohibitions only entered into application on 2 February 2025, the enforcement rules are “not yet applicable,” and there is “a lack of practical experience with the prohibitions” — so a more substantive evaluation will only be possible once the prohibitions have applied for at least a year and common challenges or regulatory gaps emerge. On high-risk systems, the report says evaluating how the rules function and spotting gaps “will be facilitated once the Commission guidelines on the classification of high-risk AI systems are published and practical experience has been acquired,” and positions AI regulatory sandboxes as a mechanism for regulatory learning and evidence collection. The report flags one substantive issue: a potential regulatory gap on AI systems generating child sexual abuse material and non-consensual intimate content, currently not prohibited by Article 5 — noting the 7 May 2026 AI Omnibus political agreement includes a ban on such systems (e.g. “nudification” apps). It states the Commission has flagged specific (unnamed in the page) AI systems for monitoring in subsequent reviews, and that a targeted consultation on the draft high-risk classification guidelines has been launched whose results will inform next year’s review.

Notable quotes

“A more substantive evaluation of the application of Article 5 AI Act will only become possible after the prohibitions have applied for at least a year and common challenges or regulatory gaps begin to emerge.” — Commission report library item, 22 May 2026

“It is also expected that regulatory sandboxes established in accordance with the AI Act will be mechanism for regulatory learning and evidence collection that will help to identify possible regulatory gaps and challenges in interpretation.” — Commission report library item, 22 May 2026

What’s speculative vs. asserted

• Asserted: the 22 May 2026 adoption of the report; its Article 112(1) basis and methodology; the conclusion that Article 5 / Annex III review is early-stage pending a year of practice and the high-risk classification guidelines; the flagged CSAM / non-consensual-intimate-content gap and its link to the 7 May 2026 AI Omnibus ban; the launch of the targeted consultation on draft high-risk classification guidelines.
• Speculative / forward-looking: which specific AI systems are “flagged for monitoring” (not named on the page; PDF not retrieved); whether and how Annex III or Article 5 will actually be amended in future reviews; the practical adequacy of sandboxes as an evidence-collection mechanism.

Topics this feeds

• EU AI Act — annual Article 112(1) review of prohibitions and high-risk scope.
• EU AI Office (European Commission) — entity tracked change.

Open questions raised

• Which specific AI systems has the Commission flagged for monitoring in subsequent reviews (named in the PDF, not retrieved)?
• How does the “review is early-stage / needs a year of practice” posture interact with the 2 August 2026 applicability cliff for other parts of the Act?

Ingestion note

Library item page fetched directly and in full; the report PDF (newsroom document 128788) was linked but not retrieved — confirm the flagged-systems detail against the PDF before client use. Post-dates the 19 May 2026 draft high-risk classification guidelines already in the wiki (S-2026-05-19-eu-ai-office-high-risk-draft); the two are complementary (the review report explicitly defers the substantive high-risk evaluation to the guidelines once published).