The AI Act is delayed but already biting
„The Digital Omnibus only extended deadlines for high-risk items; transparency and GPAI rules remain in force, and the AI Office is already requesting information from companies, so compliance cannot be postponed.”
Counter-narrative
The delay indicates that fundamental regulatory problems remain unresolved. In many member states, enforcement authorities are missing, implementation is lagging, and Big Tech is lobbying to reshape the regulation. This view is held by Big Tech players and some critical analysts.
Carriers
European Commission and AI Office, International law firms and professional analyses, consultants, Professional and news media (e.g., Politico), national AI authorities
Analysis
The narrative's twofold message is that a delay is not an exemption, so preparation must be funded now. It is fed by the enforcement phase that began on 2 August, the AI Office's information requests and the first harmonised standard. It gives consultants and law firms compliance work, and gives the Commission an argument that voluntary commitments do not replace mandatory compliance. The weak point is that the real enforcement test is still to come. Many member states lack an authority, the compliance advantage tied to the first standard is delayed, and the trilogue may bring further amendments.
What's at stake: Whether binding EU regulation can remain credible after the postponements.
Narrative sub-threads
Omnibus: only high-risk items delayed
Regulation 2026/1744 (Digital Omnibus on AI) entered into force on July 27, delaying key obligations for high-risk systems to December 2027 and August 2028; for AI-based medical devices, the deadline is August 2, 2028. This is not a general delay: transparency obligations under Article 50, AI proficiency, and GPAI oversight did not slip.
This branch is based on legislative events and the deadline calendar; the evidence is analysts' interpretation of the amendment, not regulatory action.
→ lasting · evidence: strongArticle 50 labelling requirements take sharper shape
Since August 2, the transparency rules of Article 50 apply: chatbots must disclose that they are AI, and deepfakes and many kinds of AI-generated content must be labeled. Providers are responding with watermarking, but corporate readiness is lacking: a Dutch survey found that only 14 out of 100 companies clearly label their chatbot, and according to Economx, most Hungarian companies do not yet know what to prepare for.
Corporate and provider adaptation is at the center, including watermarking by Anthropic and OpenAI and a Dutch survey showing that 14 out of 100 companies label their chatbot.
↗ rising · evidence: strongThe AI Office is already requesting information
From August 2, the Commission may act as a central supervisor against GPAI providers, imposing fines of up to €15 million or 3% of turnover. The AI Office has begun requesting information: reports indicate it sent security and compliance inquiries to more than 30 AI companies, and requested information from OpenAI and Anthropic regarding incidents involving agents. Meanwhile, many member states have not yet established the authorities responsible for enforcement.
The actor here is EU supervision, and the mechanism is the specific regulatory step: information requests, the GPAI code, and reporting channels, with a fine cap of 15 million euros or 3 percent.
↗ rising · evidence: moderateStandards and compliance presumption
The first European standard linked to the AI Act, EN 18286:2026, is ready. Published in August, it entered the EU Official Journal in October as the first harmonized standard. Upon approval, presumption of conformity was not yet attached, but since gaining harmonized status, it can provide presumption of conformity for relevant requirements.
This strand follows technical standardization, not legislation or regulatory action; delays in standards were also one of the main reasons for the postponement.
↘ fading · evidence: mediumThe Hungarian AI regulatory framework takes shape
Hungary has adopted an AI law and a government decree that allow fines of up to 5.7 billion forints for high-risk AI systems. According to the August 31 issue of the Hungarian Gazette (Magyar Közlöny), AI market surveillance was transferred to the Minister of Science and Technology, and according to Jogászvilág, the Deputy State Secretariat for Artificial Intelligence has been established.
It shows the construction of the domestic institutional and competence framework: a law, the Hungarian Gazette, a ministry organization, and the Artificial Intelligence Office, not EU-level action.
Sub-thread stories
Omnibus: only high-risk items delayed
Article 50 labelling requirements take sharper shape
The AI Office is already requesting information
Standards and compliance presumption
The Hungarian AI regulatory framework takes shape
Google search interest
Hungary
| Topic and search term | Weekly | Average |
|---|---|---|
| The AI Act is delayed but already bitingai act | 2.1 | |
| Omnibus: only high-risk items delayeddigital omnibus | 0.8 | |
| Article 50 labelling requirements take sharper shapeAI content labeling | 0 | |
| The AI Office is already requesting informationai office | 1.3 | |
| Standards and compliance presumptionAI standard | 0 | |
| The Hungarian AI regulatory framework takes shapeAI law | 3.8 |
World
| Topic and search term | Weekly | Average |
|---|---|---|
| The AI Act is delayed but already bitingeu ai act | 3 | |
| Omnibus: only high-risk items delayeddigital omnibus | 0.1 | |
| Article 50 labelling requirements take sharper shapeai transparency | 2.2 | |
| The AI Office is already requesting informationai office | 16.9 | |
| Standards and compliance presumptionharmonised standards ai | 0 | |
| The Hungarian AI regulatory framework takes shapeai market surveillance | 0.1 |
Google Trends relative search index over the period studied, weekly average. A narrative and its sub-threads appear in a single comparison: 100 is the group's most-searched day, and the figures can only be compared within the group. The search term was chosen by machine analysis; it measures how much the term is searched, not how widespread the narrative is.
Who is affected and how
- compliance specialist
- Role and risk classification, and quality assurance must be aligned with the first harmonized standard now, before high-risk rules take effect in December 2027.
- corporate lawyer
- For services built on GPAI models, documentation must be requested from the provider, as the AI Office's information requests originated at the top of the chain with large model providers.
- lawyer
- Deadlines for high-risk systems under Annex III have slipped, but transparency obligations under Article 50 and GPAI rules are already in force; this must be clearly distinguished in client communication.
Hungary
Rapid institution-building is underway in Hungary. The AI Act and government decree were adopted, with fines up to 5.7 billion forints for high-risk systems (2026-09-21-01). Market supervision moved to the Minister of Science and Technology (2026-08-31-12), and the Deputy State Secretariat for Artificial Intelligence was established (2026-09-28-10).
Weekly trend
Turning points
Related narratives in the News magazine
Supporting stories
Machine analysis based on story titles and summaries, with a verification pass. The source of the facts is the article linked at each story.