Abstract
<jats:sec> <jats:title>Context</jats:title> <jats:p>The European Union's regulatory architecture for artificial intelligence in safety-critical aviation has shifted twice between November 2024 and June 2026. The original 2 August 2026 high-risk obligations under Article 6(1) of Regulation (EU) 2024/1689 were deferred to 2 August 2027 for Annex I product-embedded systems by Article 113(3)(c); the Digital Omnibus on AI, on which Council and Parliament reached provisional agreement on 7 May 2026, would further defer that date to 2 August 2028 (Council of the EU 2026a; Gibson Dunn 2026). Throughout this period, the European Union Aviation Safety Agency has been executing its own rulemaking response under Task RMT.0742, with NPA 2025-07 published on 10 November 2025 and AI Concept Paper Issue 03 released for comment on 3 June 2026 (EASA 2025a; EASA 2026a).</jats:p> </jats:sec> <jats:sec> <jats:title>Findings</jats:title> <jats:p>Five findings drive this Dossier. First, the operative compliance regime for safety-critical aviation AI is EASA's sectoral architecture, not the AI Act's horizontal apparatus. Second, vendor compliance tooling does not address the conformity-assessment route that applies to AI safety components in type-certified aircraft. Third, the standards-readiness gap at CEN-CENELEC JTC 21 has become a substantive regulatory-policy mechanism. Fourth, the capital-pace pattern that caught earlier European eVTOL entrants — capital exhaustion against the regulator's actual rather than stated timeline — is the pattern that will catch AI-component projects whose budgets assume nominal deadlines. Fifth, sovereign capability builders with civil-defence dual portfolios are accumulating an unmatchable position during the methodology window.</jats:p> </jats:sec> <jats:sec> <jats:title>Implications</jats:title> <jats:p>Investment decisions, function-ownership decisions, and certification-pathway commitments made between June 2026 and August 2027 will determine which European aerospace and defence operators carry safety-critical AI through to type certification, and which expend capital on the wrong conformity-assessment route. The Dossier provides three proprietary instruments — the EASA-AI Sovereign Calibration Audit (ESCA), the Dual-Compliance Capital Allocation Model (DCCAM), and the Sovereign Position Scorecard (SPS) — that the purchaser uses to make those decisions operationally.</jats:p> </jats:sec> <jats:sec> <jats:title>Confidence</jats:title> <jats:p>CONFIRMED on the legal architecture and primary EASA documents; REPORTED on commercial cases, market structure, and forensic narrative; INFERRED on the deep-pattern synthesis that the methodology window favours sovereign capability builders over compliance buyers. Where principal findings rest on INFERRED reasoning, the in-text flags identify the analytical move.</jats:p> </jats:sec>