The Digital Omnibus Proposal aims to streamline the European Union’s digital regulatory framework but raises important concerns. This paper highlights risks related to reduced traceability of AI training data, weakened links between data governance and high-risk classification, and potential inconsistencies arising from simplified data access and reporting mechanisms. It argues that these changes may undermine effective risk assessment, shift complexity to downstream actors, and create legal uncertainty. To address these issues, the paper proposes targeted recommendations, including enhanced transparency and notification requirements for AI training data, safeguards to ensure that data availability does not affect risk classification, ex ante assessments for high-risk data reuse, and stronger governance and accountability measures for the centralised incident reporting system. These measures aim to preserve regulatory coherence, risk sensitivity, and the EU’s broader objectives of trustworthy and sovereign AI governance.