The headlines from the May 7 Omnibus deal are settled: the December 2, 2027 deadline for Annex III high-risk systems, SME documentation relief, the GPAI marking gap. Compliance teams know those terms. What hasn’t circulated as widely are two structural amendments buried in the provisional text – amendments that affect machinery operators and AI literacy program owners in ways the headline coverage didn’t reach.
According to Covington & Burling’s legal analysis of the provisional Omnibus text, AI-enabled machinery has reportedly been moved from Annex I Section A to Annex I Section B. That’s not a clerical change. Section A systems carry dual-compliance obligations, both the AI Act and the relevant sector regulation. Section B systems follow the sector regulation as the primary framework. If the final text confirms this reclassification, machinery manufacturers and industrial AI deployers would trade a dual-compliance model for a sector-specific one. That means one compliance track instead of two. The Machinery Regulation, not the AI Act, becomes the governing framework for conformity assessment. Worth noting: this reclassification is present in Covington’s analysis of the provisional text but hasn’t been independently corroborated by a primary EU source in . Monitor for confirmation in the Official Journal publication.
The second amendment is subtler. The AI literacy obligation, Article 4, which requires providers and deployers to ensure their staff have sufficient AI literacy, has reportedly been softened. Per the same legal analysis, the obligation now requires organizations to “take measures to support the development of” AI literacy, rather than to “ensure” it. The distinction matters operationally. “Ensure” implies a verifiable outcome standard: demonstrate your people have the required literacy, or explain why not. “Take measures to support” shifts the obligation toward process: show you’ve run training, offered resources, built the program. The specific amended language hasn’t been confirmed against the final text, so compliance teams should treat this as directionally reliable but not yet actionable at the policy-writing level.
AI-Enabled Machinery: Compliance Pathway Change (Provisional)
Verification
Partial Covington & Burling legal analysis of provisional Omnibus text (insideprivacy.com, May 18, 2026) Machinery reclassification: single T3 source, no independent T1/T2 corroboration this cycle. AI literacy language change: specific amended text not confirmed in available source excerpt. Both claims pending formal adoption and Official Journal publication.Both amendments arrive in the context of a deadline most compliance teams have already planned around. The European Commission’s official AI regulatory framework page confirms December 2, 2027 as the applicability date for Annex III high-risk systems in biometrics, critical infrastructure, education, employment, migration, and asylum. The August 2, 2026 date remains in place for GPAI model obligations and transparency and labeling requirements.
The catch is that provisional doesn’t mean final. The Omnibus agreement must still be formally adopted and published in the Official Journal of the European Union. Technical amendments can shift between provisional agreement and formal publication. Compliance teams reviewing the machinery reclassification or the literacy softening should build their planning scenarios on confirmed text, not provisional analysis.
Warning
The Omnibus provisional agreement is not yet formal law. Technical amendments can change between provisional agreement and Official Journal publication. Compliance planning based on these amendments should be treated as conditional scenarios until the final text is published. Monitor the Official Journal of the European Union for formal adoption.
Don’t expect the AI Office to fill the literacy gap quickly. Even if the obligation is softened in the final text, guidance on what “take measures to support” requires in practice is unlikely before Q3 2026 at the earliest, and likely later. The operational question of what constitutes sufficient AI literacy programming under a process standard, rather than an outcome standard, is exactly the kind of implementation detail that takes months to specify. HR and L&D teams running AI literacy programs should document their methodology now, regardless of the final language.
The real question is whether machinery operators who’ve been planning for dual-compliance exposure will revise their conformity assessment roadmaps on the strength of provisional analysis alone. The directional signal from the Omnibus is clear: sector-specific pathways are the preferred model. But acting on an amendment that hasn’t cleared formal adoption carries its own compliance risk. The safer path is a conditional planning scenario, map the Section B pathway now so you’re ready to execute when the text is confirmed, while holding the Section A dual-compliance track as your active obligation.