EU AI Act delay creates gap
Delays in the EU’s AI Act are creating a regulatory window where some high‑risk systems could avoid immediate oversight, forcing companies to self‑impose stricter compliance if they want predictable risk posture. That regulatory uncertainty should be surfaced in product and risk reviews now. (techpolicy.press)
On 26 March 2026 the European Parliament adopted its position on the AI Act omnibus package by 569 votes to 45 with 23 abstentions. (aiactblog.nl) That parliamentary position fixes compliance deadlines for Annex III “high‑risk” AI use cases to 2 December 2027, replacing the Commission’s original enforcement trigger of 2 August 2026 for those systems. (biometricupdate.com) Annex III explicitly lists sectors such as remote biometric identification, critical infrastructure, education, employment (including hiring algorithms), essential services, law enforcement, migration and justice as high‑risk categories. (artificialintelligenceact.eu) The European Commission proposed the “Digital Omnibus” simplification package in November 2025 to align AI Act enforcement with the completion of harmonised technical standards and guidance. (techpolicy.press) IMCO and LIBE committee votes in mid‑March 2026 backed fixed 2027 deadlines and the Council published its position on 13 March 2026 ahead of the plenary mandate. (aiacto.eu) Lobbying from trade groups including the Computer and Communications Industry Association and a public letter from 56 EU‑based AI firms were cited in Brussels reporting as part of the pressure for simplification. (techpolicy.press) Consultancy and CIO coverage urged organisations to “plan as if” high‑risk obligations might still apply on the earlier timetable, while conformity assessment rules differentiate internal (Annex VII) self‑assessment from notified‑body third‑party assessment for certain use cases such as remote biometrics. (computerworld.com) Leadership briefings that map each product to its Annex III use‑case, state the alternative enforcement dates (2 August 2026 vs 2 December 2027) and flag the likely conformity‑assessment route convert political uncertainty into decision‑grade inputs. (artificialintelligenceact.eu) Non‑compliance exposure should be quantified in the same update using the Act’s penalty tiers—administrative fines reach up to €35 million or 7% of global turnover for prohibited practices and up to €20 million or 4% for specified high‑risk breaches under the Act’s enforcement provisions. (ai-act-service-desk.ec.europa.eu) Analysts forecast rising spend on AI governance—Gartner projects roughly $492 million in AI‑governance spending in 2026 with growth toward $1 billion by 2030—making early compliance investment a measurable budget line for boards weighing immediate controls versus a regulatory delay. (gartner.com) Parliament’s mandate will head into trilogue negotiations with the Council, so a recorded board posture that commits to either immediate high‑risk controls or an explicit acceptance of the regulatory gap will transform the shifting political timetable into a concrete procurement and certification plan. (aiactblog.nl)
