FrenchIT Support
European Regulation · update July 2026
The EU AI Act: what really changes for your business
With deferred deadlines, existing obligations, and debates about its cost for SMEs, the AI Act remains poorly understood. Here is an up-to-date and balanced breakdown – advantages, legitimate criticisms, and what really needs to be done today.
Updated calendar (rescheduling of the Digital Omnibus) Advantages and criticisms, without bias GVISION Compliance Support
Feb 2025Prohibitions and AI literacy — already in force
August 2025General-purpose AI obligations — already in force
December 2027New high-risk deadline (trilogue agreement, May 2026)
€35 millionmaximum fine for prohibited practices
What is the European AI Act?
The AI Act (European regulation on artificial intelligence) classifies AI systems according to their risk level — unacceptable, high, limited, or minimal — and imposes obligations proportionate to each category, ranging from outright bans to simple transparency requirements. Coming into force in August 2024, it will be applied in stages until 2027-2028, with a recently revised timetable to allow more time for businesses and regulators.
What has just changed
The timeline for the AI Act has been thoroughly revised.
The most important news of 2026: the heaviest obligations of the AI Act have been postponed, due to a lack of ready tools to enforce them.
News — May 2026
A trilogue agreement postpones «high-risk» obligations until December 2027
Given the absence of harmonised technical standards and operational certification bodies, the European Commission proposed a «Digital Omnibus» package at the end of 2025 to synchronise the AI Act's timeline with on-the-ground reality. The European Parliament approved it on 26 March 2026, and a trilogue agreement was reached on 6 May 2026. Its formal adoption and publication in the Official Journal are expected during the summer of 2026.
What this means in practice: the prohibitions and the obligation of AI literacy remain fully applicable since February 2025. It is the heaviest obligations — technical documentation, conformity assessment by a third party for high-risk systems.
Detailed calendar
Initial deadlines vs revised deadlines
| Obligation | Initial due date | Revised deadline | Status |
|---|---|---|---|
| Prohibited Practices & AI Literacy | 2 February 2025 | Unchanged | In force |
| General-Purpose AI Obligations (GPAI) | 2 August 2025 | Unchanged | In force |
| High-risk systems (Annex III) | 2 August 2026 | 2 December 2027 | Trilogue agreement, formal adoption expected summer 2026 |
| AI integrated into regulated products (Annex I) | 2 August 2027 | 2 August 2028 | Trilogue agreement |
| Transparency & labelling of AI content (Article 50) | 2 August 2026 | Late 2026 / Early 2027 | Postponement of 3 to 6 months |
Subject to formal adoption and publication in the Official Journal of the EU, expected in summer 2026. Until it is published, the current legal deadline technically remains 2 August 2026 — hence the benefit of not waiting to prepare.
A two-voice debate
What the AI Act brings, what it costs
Neither pure opportunity nor pure constraint: here are the arguments from both sides of the debate, as they are documented today.
What the supporters advocate
A trusted advantageDemonstrating ethical and transparent AI is becoming a commercial argument, especially in sensitive sectors (healthcare, finance, HR).
A framework that can stimulate innovationRegulatory sandboxes allow for the testing of high-risk systems under supervision, with clear rules from the outset.
Increased investor confidenceA stable framework directs capital towards compliant companies in an emerging responsible AI market.
A precursor advantageThe AI Act could become a benchmark for other jurisdictions — complying with it early limits the work needed later.
What critics are pointing out
An excessive cost for SMEsUp to 20 % of an SME’s R&D budget is spent on ensuring compliance for a single high-risk system, compared with proportionally much lower expenditure for large companies.
A heavy documentary burdenSome estimates suggest that up to 30 % of technical capacity is devoted solely to compliance documentation.
A persistent legal uncertaintyThe delay in harmonised technical standards and certification bodies has directly caused the rescheduling of the timetable itself.
A risk of «compliance debt»MEPs warn of the risk that companies could deploy high-risk systems before December 2027 without a third-party audit, before having to urgently catch up on compliance.
Sources: Implement Consulting Group, AI Policy Bulletin, SQ Magazine, A&O Shearman, WEnvision — analyses of the Digital Omnibus 2026 and its impacts.
Orders of magnitude
The cost of compliance, in concrete terms
~52 000 €average annual compliance cost for a high-risk AI system
£50,000–£500,000cost of compliance range for an SME, by complexity
10–15 %share of the AI budget allocated to compliance each year
1–2 ETPemployees often dedicated to AI compliance in a startup
UK: SQ Magazine — EU AI Act Compliance Cost Statistics 2026. Indicative orders of magnitude, variable according to sector and the company's AI maturity.
Make no mistake
What already applies, irrespective of the deferral
The report only concerns the most burdensome obligations. Several rules have been active since February 2025.
Prohibited practices
Subliminal manipulation, exploitation of vulnerabilities, social scoring, certain uses of biometric recognition.
AI Literacy for Staff
Any business using AI must ensure a sufficient level of understanding within its relevant teams.
General-purpose AI obligations
Providers of generative artificial intelligence (GPAI) must comply with transparency and risk management obligations from August 2025.
And in Belgium?
Several Belgian authorities share the implementation of the AI Act: the FPS Economy for market surveillance, and the CSA (Conseil Supérieur de l’Audiovisuel) designated as the authority responsible for the protection of fundamental rights. This division between authorities reinforces the benefit of seeking assistance to know precisely who to contact according to your situation.
Discover our ISO, NIS2 & GDPR compliance supportOur recommendation
Whatever happens politically, preparing yourself remains a winning strategy.
The debate about the timetable changes nothing about one reality: the companies that map and document their uses of AI from now on avoid rushing, whatever the final outcome of the Digital Omnibus.
Map
Identify all AI systems used or developed within your company.
Classifier
Determine the risk level of each system according to the AI Act criteria.
Train
Implement mandatory AI literacy for your relevant teams.
Document
Prepare the expected technical documentation for high-risk systems.
Monitor
Maintaining continuous regulatory monitoring in the face of an ever-changing timetable.
Deploying AI while remaining compliant
GVISION also supports your artificial intelligence projects — Copilot, AI agents, chatbots, automations — with compliance built-in from the design stage rather than added as an afterthought.
Discover our AI solutions for businesses3 100+clients supported
6 400+servers monitored
1 500+devices managed
200+Years of combined expertise
Questions fréquentes
Everything you need to know about the AI Act in 2026
What is the European AI Act?
The AI Act is the European regulation governing the development and use of artificial intelligence. It classifies AI systems according to their risk level (unacceptable, high, limited, minimal) and imposes proportionate obligations, ranging from outright prohibition to simple transparency requirements.
Is the deadline of 2 August 2026 cancelled?
A political agreement (trilogue) was reached in May 2026 to postpone the deadlines for high-risk systems until December 2027. Its formal adoption and publication in the Official Journal of the EU are expected in summer 2026. Until it is officially published, the legal deadline technically remains 2 August 2026.
Which obligations of the AI Act are already in force today?
From 2 February 2025, AI practices deemed unacceptable are banned (manipulation, exploitation of vulnerabilities, social scoring), and companies must ensure a level of AI literacy among their staff. From August 2025, providers of general-purpose AI models will have specific obligations.
Is my company affected by the AI Act?
Any company developing, deploying, or using an AI system within the European Union is affected, to varying degrees depending on the risk level of the system. Even SMEs using everyday AI tools must comply with the AI literacy obligation for their teams.
Quels sont les principaux avantages de l’AI Act pour les entreprises ?
Compliance can build customer and partner trust, offer a first-mover advantage if other jurisdictions align with this model, facilitate access to investors who value responsible AI, and clarify the framework within which to innovate via regulatory sandboxes.
Les principales critiques adressées à l'AI Act portent sur le fait qu'il pourrait étouffer l'innovation, en particulier pour les petites et moyennes entreprises (PME), en raison des exigences de conformité complexes et coûteuses. Certains craignent également que la législation ne soit pas suffisamment précise ou assez rapidement adaptable aux avancées rapides de l'IA. D'autres critiques portent sur l'équilibre entre la nécessité de réglementer l'IA pour des raisons éthiques et de sécurité, et le risque de surréglementation qui pourrait nuire à la compétitivité de l'Europe. Enfin, il y a des préoccupations quant à la manière dont l'application et le contrôle seront mis en œuvre efficacement.
The main criticisms centre on the cost of compliance, which can amount to as much as 20 % of an SME’s R&D budget for a single high-risk system; the disproportionate administrative burden on small organisations; and the legal uncertainty arising from delays in the publication of technical standards.
What are the risks for a company that does not comply with the AI Act?
Fines can reach €35 million or 7 % of global turnover for prohibited practices, and lower but still significant amounts for other breaches. In addition to financial penalties, non-compliance exposes organisations to reputational and contractual risks.
How can GVISION help me comply with the AI Act?
GVISION maps your AI systems, assesses their risk level, implements mandatory AI literacy for your teams, prepares the required technical documentation, and ensures continuous regulatory monitoring, in line with your obligations. GDPR and NIS2.
Give an update on your exposure to the AI Act.
Our experts help you distinguish what is already mandatory from what remains uncertain — and to act on both without panicking.