Guide AI Compliance

EU AI Act: What Is Now in Force and What the Digital Omnibus Changed

The EU AI Act (Regulation (EU) 2024/1689) already applies in part. Prohibitions and the AI literacy duty have applied since February 2025, and Article 50 transparency duties since August 2026. The Digital Omnibus moved Annex III high-risk obligations to December 2027 and Annex I obligations to August 2028, and added two prohibitions from December 2026. This guide sets out what applies now, what changed, and what to prepare.

4 August 2026
Updated 10 November 2026
7 min read
Key takeaways
  • The Digital Omnibus (Regulation (EU) 2026/1744, in force since 27 July 2026) moved Annex III high-risk obligations to 2 December 2027 and Annex I product-related high-risk obligations to 2 August 2028
  • The same amendment adds two prohibited practices from 2 December 2026: AI that produces non-consensual intimate images of identifiable people, and AI that produces child sexual abuse material
  • Article 50 transparency duties have applied since 2 August 2026, and they differ by role: providers mark synthetic output, deployers disclose deepfakes and certain public-interest text
  • Being listed in Annex III does not settle classification. Article 6(3) lets some narrow, preparatory systems fall outside high risk, but systems that profile people stay high-risk
  • Using ChatGPT, Copilot or Gemini does not pass the model provider's compliance on to you. You assess your own duties for each use case
  • AI literacy under Article 4 has applied since February 2025 to providers and deployers, whatever the risk tier
  • Fines reach €35 million or 7% of worldwide turnover for prohibited practices

What the EU AI Act is

The EU AI Act (Regulation (EU) 2024/1689) is the first comprehensive legal framework for artificial intelligence. It was published on 12 July 2024 and entered into force on 1 August 2024. It applies across sectors, within the scope rules and exclusions set out in Article 2, and its obligations depend on the role you play and the risk of the system.

The Act distinguishes two main actors. Providers develop an AI system, or have one developed, and place it on the market or put it into service under their own name: software vendors, SaaS companies, AI tool developers. Deployers use an AI system under their own authority in a professional context. Most businesses using off-the-shelf AI tools for HR, customer service or operations are deployers. Providers carry the heavier obligations, but deployers of high-risk AI also have duties of their own.

Scope

The Act applies to providers that place AI systems or general-purpose AI models on the EU market, wherever they are established; to deployers established or located in the EU; and to providers and deployers in third countries where the output of their AI system is used in the EU. The UK is outside the EU and has no equivalence arrangement under the Act, so a UK business comes into scope only through these tests: for example, by selling an AI system into the EU, deploying AI through an EU entity, or producing AI output that is used in the EU. Assess each party, system and activity separately.

The four risk tiers

The four tiers below are a useful way to understand the Act, but they are a simplification. Article 50 transparency duties can apply whether or not a system is high-risk, general-purpose AI models have their own regime, and the Article 4 AI literacy duty applies to providers and deployers of any AI system.

Unacceptable risk, prohibited Since 2 February 2025, with two additions from 2 December 2026
Practices listed in Article 5 are prohibited, subject to the specific exceptions written into the Article.
Harmful social scoring Subliminal manipulation Emotion recognition at work or in education Facial recognition database scraping Non-consensual intimate imagery (from Dec 2026)
High risk Annex III from 2 December 2027, Annex I from 2 August 2028
AI systems used in the Annex III areas, and AI that is a safety component of products covered by the EU harmonisation legislation in Annex I. Both providers and deployers have obligations.
CV screening and recruitment AI Employee performance monitoring Creditworthiness assessment of individuals Student assessment AI AI safety components in regulated products
Transparency duties Article 50, since 2 August 2026
Systems that interact with people, generate synthetic content, or perform emotion recognition or biometric categorisation. The duty depends on whether you are the provider or the deployer (see Article 50 transparency).
Customer service chatbots Image, audio and video generators Deepfake content
Minimal risk No risk-specific obligations
Most AI applications, such as spam filters, recommendation engines and most productivity automation, fall here. There are no obligations specific to this tier, but AI literacy still applies, and providers can adopt voluntary codes of conduct.
Spam filters AI-powered search Inventory forecasting Grammar correction

The detailed obligations

The detailed obligations are in two companion articles.

Prohibited AI practices and high-risk AI systems covers the practices Article 5 bans, the two routes into the high-risk category under Annex III and Annex I, when an Annex III system is not high risk, and what providers and deployers of high-risk systems must do.

Article 50 transparency, GPAI models and fines covers the transparency duties for providers and deployers, what the obligations on general-purpose AI model providers mean for organisations that use tools such as ChatGPT or Microsoft 365 Copilot, and the fine structure.

The implementation timeline

1 Aug 2024
Act enters into force
Regulation (EU) 2024/1689 published and in force. Transition periods begin.
2 Feb 2025
Prohibitions and AI literacy apply
General provisions, the Article 4 AI literacy duty and the original Article 5 prohibited practices apply.
2 Aug 2025
GPAI model obligations apply
Providers of general-purpose AI models must meet documentation, transparency and copyright requirements. Providers of models with systemic risk face additional duties. Governance rules and penalties provisions also apply.
2 Aug 2026
Article 50 transparency obligations apply
Chatbot disclosure, machine-readable marking of synthetic content and deepfake disclosure apply, with duties split between providers and deployers. The Commission's enforcement powers over GPAI providers also apply.
2 Dec 2026
New prohibitions and marking deadline
The two prohibitions added by the Digital Omnibus apply. Generative systems placed on the market before 2 August 2026 must have machine-readable marking in place.
2 Dec 2027
Annex III high-risk obligations apply
Moved from August 2026 by the Digital Omnibus. Provider and deployer obligations apply to high-risk systems in the Annex III areas, such as employment, credit, education and biometrics.
2 Aug 2028
Annex I high-risk obligations apply
Moved from August 2027 by the Digital Omnibus. Requirements apply to high-risk AI systems linked to products covered by the EU harmonisation legislation in Annex I (Article 6(1)), subject to the transitional rules.

For the full list of dates and transitional rules, see our EU AI Act enforcement timeline.

What most businesses need to do now

The December 2027 date gives organisations time for a structured process, but some duties already apply. Start with the obligations in force, then build the inventory you need for the high-risk deadlines.

Step 1: Put AI literacy measures in place

Article 4 has applied since 2 February 2025. After the Digital Omnibus, providers and deployers must take measures to support AI literacy among their staff and others who operate or use AI systems on their behalf, taking into account their knowledge, the context of use and the people affected. There is no single required course or certificate, and the Act does not require you to guarantee a particular level for each person. Proportionate measures that you can evidence are what count.

Step 2: Inventory your AI use

List every AI system in use across your organisation, including tools individual teams have adopted without IT review. Include productivity AI such as Copilot and ChatGPT, HR platforms with AI features, customer-facing chatbots, fraud detection tools and analytics platforms that inform decisions about people. For AI that can take actions on behalf of users, our guide to agentic AI security risks covers the specific controls.

Step 3: Classify each system and your role

For each system, record whether you are the provider or the deployer, then the tier. The key question for most SMEs: does the system make or materially assist decisions about people in employment, education, credit, insurance or essential services? If so, check it against Annex III and the Article 6(3) conditions. Then check separately whether Article 50 applies.

Step 4: Check your HR and recruitment AI

Employment is the Annex III area that touches the widest range of organisations. If a tool screens CVs, ranks candidates, monitors productivity or allocates tasks, ask the vendor directly whether it treats the system as high-risk under the EU AI Act, what its compliance plan is for December 2027, and what documentation it will give you.

Step 5: Review your prohibited practice exposure

Check each Article 5 prohibition against your current tools. Emotion recognition at work is the most common exposure for private organisations. If you provide image or video generation tools, review the December 2026 prohibitions now.

Step 6: Update policies and inform your workforce

Update your acceptable use policy to cover generative AI, record which tools are approved and for what purposes, and confirm that the people responsible for AI oversight have the authority and time to exercise it. Where a high-risk system will be used at work, plan how you will inform workers and their representatives before it goes live.

Our audits and compliance team runs AI Act readiness assessments, from inventory to gap analysis to remediation. Our AI advisory practice supports businesses deploying AI in regulated sectors, including financial services, healthcare and hospitality.


Frequently asked questions

Does the EU AI Act apply to businesses outside the EU?

It can. The Act covers providers that place AI systems or GPAI models on the EU market, wherever they are established, deployers established or located in the EU, and providers and deployers in third countries where the output of their AI system is used in the EU. Whether a particular business is covered depends on its role, the system, the activity and the exclusions in Article 2, so assess each case against those tests.

What is the difference between a provider and a deployer under the EU AI Act?

A provider develops an AI system, or has one developed, and places it on the market or puts it into service under its own name. A deployer uses an AI system under its own authority in a professional context. Most businesses using tools such as ChatGPT, Microsoft 365 Copilot or third-party recruitment AI are deployers. Providers carry heavier obligations, such as technical documentation and conformity assessment. Deployers of high-risk AI have their own duties under Article 26.

Sources

Ada Bicker
About the author
Cybersecurity Consultant, Cyvra

Ada has more than five years' experience in information security, ISO 27001, SOC 2 and EU compliance, and writes about DORA, the EU AI Act and NIS2. View full profile

EU AI Act Readiness

Not sure where your AI use lands?

We help organisations map their AI inventory, identify high-risk exposure, and build a compliance plan ahead of the December 2027 Annex III deadline.