The tech industry’s read on the European Union’s newly concluded AI Act “omnibus” simplification deal: better, but not done. On May 7, 2026, the Information Technology Industry Council (ITI) — a global tech trade association — published a statement from Guido Lobrano, its Director General for Europe, welcoming the deal reached in the AI Act’s omnibus trilogues while pressing EU policymakers to keep going. This is coverage of the EU AI Act‘s own simplification track — a separate development from the US state-law landscape (California’s SB 53, Colorado’s AI Act) that CASRAI covers elsewhere.
What ITI says the deal gets right
According to ITI’s May 7 press release, the conclusion of the AI Omnibus trilogues delivers two changes ITI’s membership had been pushing for: a delay to key high-risk AI system compliance obligations, and streamlined rules aimed at reducing duplicative regulation for industrial AI. ITI did not publish the specific revised compliance dates in this release, and CASRAI has not independently verified them against the AI Act’s own text as of this writing — this piece works from what ITI itself has said, not from the underlying legislative text.
Lobrano’s statement credits both changes as substantive, not cosmetic:
“The delay on high-risk requirements and the move to streamline rules for industrial AI are welcome and necessary steps.”
What ITI says is still missing
The same statement immediately qualifies that praise. Two specific concerns carry over into the industry’s ongoing ask of EU policymakers: compliance timelines for watermarking and labelling requirements that ITI calls too tight, and continued overlap between the AI Act’s horizontal, general-purpose rules and its vertical, sector-specific ones — a friction point the industrial-AI carve-out addresses only for industrial AI, not more broadly. In full, Lobrano’s statement reads:
“The delay on high-risk requirements and the move to streamline rules for industrial AI are welcome and necessary steps. But the agreement does not fully address important challenges for the industry – from tight deadlines for watermarking and labelling requirements to continued regulatory overlaps across horizontal and vertical frameworks, beyond industrial AI. EU policymakers must continue backing an ambitious simplification agenda going forward in support of innovation and technology uptake goals.”
Read against CASRAI’s own comparison of ten frontier-AI jurisdictions, the EU AI Act remains the binding, centrally-enforced end of that spectrum — which is precisely why an industry trade group is treating a trilogue-level simplification deal as newsworthy in its own right, months before any of it takes practical effect on a compliance calendar.
CASRAI editorial note: how this lands on NIKOLAI
CASRAI’s own NIKOLAI project — an independent, unendorsed reference vocabulary for frontier-AI-safety terminology, not a standard adopted or endorsed by any lab, regulator, or trade body — does not currently have an element built specifically for watermarking or AI-generated-content labelling, the exact requirement Lobrano’s statement flags as too tight. CASRAI checked NIKOLAI’s full 64-element inventory, including a direct search across every published element for “watermark,” “label,” “disclosure,” “synthetic,” and “provenance,” before writing this section, and none returned a match. The closest existing element sits in NIKOLAI’s Transparency and Review track (N8): Evaluator Independence and Conflict of Interest, which covers who is allowed to verify a developer’s disclosures, not what a developer must disclose about AI-generated output itself. That is a related but distinct question from watermarking, so CASRAI is not forcing a crosswalk row here. This is a genuine gap in NIKOLAI’s current element set, not a shadow mapping CASRAI is choosing to omit — and it is one CASRAI is noting as a candidate area for a future element, precisely because industry statements like this one keep surfacing watermarking and labelling as a live compliance friction point.
