-3.jpg?width=784&height=1168&name=image%20(60)-3.jpg)
Digital safety
Ethics
EU AI Act: Delay of AI Rules and New Obligations 2026–2027
Mary, NexSynaptic Founder
Why the delay of the EU AI Act matters for the European AI ecosystem
updated article
The European Union is entering a new phase of artificial intelligence regulation after the European Parliament voted to delay the implementation of key parts of the EU AI Act, while final approval from the Council of the EU is still pending.
This delay, which shifts the application of rules for high‑risk AI systems to December 2027, marks a significant shift in the regulatory timeline. At the same time, the EU is introducing new safety and transparency measures, including mandatory labeling of AI‑generated content and new rules targeting deepfake technologies.
What is behind the EU’s decision to delay AI rules and soften regulation
The delay in the EU AI Act is the result of a combination of industry pressure, regulatory complexity, and the need to harmonize overlapping digital laws. European companies, from tech startups to large industrial players have warned that overlapping regulations such as the GDPR, DSA, DMA, and the AI Act place them at a competitive disadvantage compared to companies in the United States and Asia.
The European Commission recognized that rapid implementation could slow innovation, increase compliance costs, and push technological development out of Europe. As a result, the timeline is being adjusted, but the regulatory ambition remains intact.
New operational developments for 2026–2027
As the implementation timeline shifts, the European Commission has confirmed several operational measures that will shape how organizations prepare for compliance.
Technical standards for high‑risk AI systems will be released in phases, with the initial set expected by late 2026 and the full framework arriving throughout 2027. This means companies will need to begin aligning their systems before all standards are finalized, increasing the importance of internal risk governance and documentation.
Member States will expand national supervisory authorities during 2026 and 2027, creating dedicated units responsible for AI inspections, dataset reviews, risk evaluations, and compliance checks. This marks the first time the EU will introduce operational oversight of AI systems similar to regulatory supervision in sectors such as finance and healthcare.
Additionally, the Commission will launch a centralized EU incident‑reporting portal in 2026. Organizations will be required to report serious AI‑related incidents, including unexpected model behavior, failures of safety mechanisms, and harm caused by AI systems. The portal will later integrate with the EU database of high‑risk systems, establishing a new level of transparency across the AI ecosystem.
Finally, while high‑risk obligations are delayed, requirements for generative AI remain on an accelerated timeline. Measures such as content labeling, copyright safeguards, and misuse prevention will apply earlier, reinforcing the EU’s focus on transparency and user protection.
Pratical steps how to prepare for EU AI Act
1. Map AI systems across the organization
Identify all AI systems currently in use, including internal tools, vendor solutions, and experimental models. Categorize them by risk level and intended purpose.
2. Establish an AI governance team
Create a cross‑functional group responsible for compliance, including legal, IT, security, and data teams. Assign clear roles for monitoring, documentation, and incident reporting.
3. Implement continuous risk assessment
Develop a process for evaluating model behavior, data quality, and potential harm. Risk assessments should be updated regularly, not only during deployment.
4. Prepare documentation templates
Set up standardized templates for technical documentation, dataset descriptions, human oversight procedures, and model evaluation reports.
5. Introduce human‑in‑the‑loop controls
Ensure that high‑impact decisions made by AI systems include human review, override options, and clear accountability.
6. Monitor regulatory updates
Track new technical standards, supervisory authority guidelines, and EU incident‑reporting requirements as they are released throughout 2026–2027.
Checklist: What companies must have ready before 2027
- Inventory of all AI systems
- Risk classification for each system
- Documentation for training data sources
- Model evaluation and testing logs
- Human oversight procedures
- Incident‑reporting workflow
- Security controls for model misuse
- Transparency measures for generative AI
- Internal AI policy and employee training
- Vendor compliance verification
Common mistakes companies make when preparing for AI regulation
Underestimating documentation requirements
Companies assume documentation is a one‑time task, but the EU AI Act requires continuous updates.
Ignoring vendor risk
Companies often rely on third‑party AI tools without verifying whether vendors meet EU compliance standards.
Focusing only on high‑risk systems
Generative AI obligations apply earlier and to a much wider range of tools than expected.
Lack of internal training
Employees frequently use AI tools without understanding compliance implications, increasing organizational risk.
No incident‑response plan
Without a clear workflow, companies may fail to report incidents within required timeframes.
How EU AI Act affects SMEs differently than large enterprises
Small and medium‑sized enterprises face unique challenges under the EU AI Act. Limited resources make it harder to implement continuous monitoring, documentation, and governance structures. However, SMEs may benefit from simplified compliance pathways and reduced administrative burdens introduced through Omnibus VII. Larger enterprises, on the other hand, will face stricter oversight, more frequent inspections, and higher expectations for technical controls. This difference in regulatory pressure will shape how companies allocate resources and prioritize AI development in the coming years.
Expected timeline milestones (2026–2027)
- Q2 2026: Initial supervisory authority expansion
- Q3 2026: Launch of EU incident‑reporting portal
- Q4 2026: First batch of technical standards released
- Early 2027: Additional standards for high‑risk systems
- Mid 2027: Integration of incident portal with EU database
- December 2027: Full enforcement of high‑risk AI obligations
How the delay of high‑risk AI rules affects the market and industry
The most significant change is the postponement of the high‑risk AI rules to December 2027, without a specified day.
The original deadline was August 2026, meaning the delay is roughly sixteen months.
This gives industry additional time to comply with complex requirements involving strict standards for safety, transparency, data governance, human oversight, and technical documentation.
Companies now have more room to build AI governance structures, prepare risk assessments, and implement technical controls required for compliance with the EU AI Act.
Why the question of exempting industrial machinery remains unclear
Earlier interpretations suggested that industrial machinery might be exempt from the EU AI Act, but available sources do not confirm such an exemption. While some industrial systems may fall under other regulations such as the Machinery Regulation, there is no explicit confirmation in preliminary communications. Therefore, this claim cannot be considered official until the consolidated legal text is published.
How the EU AI Act tightens rules for generative AI and deepfake content
Despite delaying some obligations, the EU is introducing new restrictions targeting generative artificial intelligence.
Deepfake regulation is a central focus, but details on the ban of sexualized deepfakes are not yet finalized. According to available information, the ban will primarily target applications that generate sexualized images without consent, including so‑called “nudify” tools.
However, systems with built‑in safety mechanisms may not fall under the ban. Since the exact date of enforcement has not been officially confirmed, cautious wording is necessary.
Why watermarking AI content is essential for transparency and combating disinformation
Mandatory labeling of AI‑generated content using watermarking will take effect in November 2026, which is earlier than previously stated. This measure aims to increase transparency and reduce the risk of disinformation, manipulation, and fake news.
It is particularly important in political campaigns, media environments, and social networks, where generative AI is increasingly used to create convincing but false content. Watermarking will help users, regulators, and platforms more easily identify AI‑generated material.
How the delay of the EU AI Act fits into the broader digital regulatory strategy and the Omnibus VII initiative
The delay is part of the European Commission’s broader initiative to simplify digital regulation, known as Omnibus VII. The goal is to reduce regulatory burdens, especially for small and medium‑sized enterprises, and to strengthen the competitiveness of the European digital sector.
Companies have long warned that overlapping regulations create a complex and costly compliance environment. The Commission recognized the need to adjust the implementation pace to avoid negative impacts on innovation and investment.
Is the delay of the EU AI Act a concession to Big Tech
Although some critics argue that the EU is yielding to pressure from major technology companies, the situation is more nuanced. The delay and administrative simplification do benefit industry, but the EU is simultaneously introducing stricter measures for generative AI, particularly in protecting fundamental rights and preventing misuse. This demonstrates that the EU is not abandoning its ambition to lead in safe and ethical AI regulation, but is instead seeking a balance between protecting citizens and supporting innovation.
How the delay of the EU AI Act will affect companies, developers and the AI market
For companies and developers, the delay provides short‑term relief, but long‑term obligations remain equally demanding. Organizations will still need to conduct risk assessments, ensure data quality, maintain technical documentation, implement human oversight, and prepare for the registration of high‑risk systems in the EU database.
Generative AI will face additional obligations, including watermarking, deepfake detection, and preventing the generation of non‑consensual sexual content. All of this requires investment in AI governance, safety mechanisms, and technical infrastructure.
How the delay of AI rules affects safety, fundamental rights, and user trust
The changes have both positive and negative implications. On the positive side, the EU is introducing clear measures to combat deepfakes and sexualized content, protecting women, children, and vulnerable groups. Transparency of AI‑generated content increases user trust and reduces manipulation risks. However, delaying high‑risk AI rules means that areas such as biometrics, law enforcement AI, healthcare technologies, and critical infrastructure will remain less regulated until 2027, raising concerns among human rights organizations.
The delay changes the timeline but not the Eu AI regulation
The preliminary agreement to soften parts of the EU AI Act is a political compromise, not a retreat. Europe is trying to balance two goals: protecting fundamental rights and fostering innovation. The AI Act remains the strictest AI regulatory framework in the world, but its implementation timeline is being adjusted to ensure it is practical and sustainable. This approach gives industry time to prepare, regulators time to finalize technical standards, and society a path toward a safer and more transparent AI ecosystem.
What next?
Organizations will need to adapt their planning to a more dynamic compliance environment. The phased release of technical standards, the expansion of supervisory authorities, and the introduction of incident‑reporting obligations mean that preparation can no longer rely solely on static documentation or one‑time assessments. Instead, companies will need continuous monitoring processes, updated governance frameworks, and flexible technical controls that can adjust as new requirements are published. This shift toward ongoing compliance reflects the EU’s broader intention to ensure that AI systems remain safe, transparent, and accountable.
AI Transparency: This article was written by the author. AI tools were used to support editing and grammar refinement. This article contains AI‑generated images. The final version was reviewed by a human.
Explore related pillars
👉Digital Safety Guide - A framework for privacy and safe digital ecosystems.
👉Neurotechnology Guide - A framework for neural data, BCI technologies, and neuromorphic computing.
Browse all Ethics articles👉 ethics
Read more about AI Ethics & Regulation👇
🔙 Return to the beginning of the journey
Explore more topics:
Ethics • AI Trends • Neurotechnology • Digital Safety • Brain Science • AI Tools • Technology
