The EU AI Act is no longer a future regulation. It is being enforced in phases, and in 2026 the EU changed several of its key deadlines through the Digital Omnibus on AI. Many guides published in 2024 and 2025 now show dates that no longer apply.

If your company builds or uses AI that reaches users in the European Union, EU AI Act compliance is now a product and engineering priority, not just a legal one. This guide explains who the law applies to, what changed with the Digital Omnibus, the updated timeline through 2028, what providers and deployers must do, and a practical checklist you can start using today.

Key Takeaways

  • The EU AI Act applies to any company whose AI is used in the EU, including businesses based in India, the US, and elsewhere.
  • AI systems are regulated by risk level: unacceptable (banned), high, limited, and minimal.
  • The Digital Omnibus moved high-risk AI deadlines to 2 December 2027 (Annex III) and 2 August 2028 (Annex I).
  • Most transparency obligations and general-purpose AI rules were not delayed, so this is not a pause on regulation.
  • Fines can reach €35 million or 7% of global annual turnover for prohibited AI practices.
  • AI agents are classified by use case, so an agent making hiring or lending decisions is treated as high-risk.
  • Starting with an AI inventory and risk classification is the fastest way to begin EU AI Act compliance.

What Is the EU AI Act and Who Does It Apply To?

The EU AI Act (Regulation (EU) 2024/1689) is the world’s first comprehensive law regulating artificial intelligence. It sets rules based on how much risk an AI system poses to people’s health, safety, and fundamental rights.

The most important point for non-EU businesses is its reach. The Act applies based on where AI is used, not where the company is based. It covers:

  • Providers: companies that develop an AI system or model and place it on the EU market under their own name.
  • Deployers: organizations that use an AI system in their professional activities.
  • Importers and distributors: businesses that bring AI systems into the EU market or make them available there.

This means an AI company in India or the US that sells a SaaS product to EU customers, or whose AI output is used in the EU, may need to comply.

Also Read: HIPAA Compliance for Healthcare Software and AI

The Four Risk Categories Under the EU AI Act

Risk Categories Under the EU AI Act

The EU AI Act uses a risk-based approach. The higher the risk, the stricter the rules.

1. Unacceptable Risk (Prohibited)

These AI practices are banned outright. Examples include social scoring of individuals, manipulative techniques that exploit vulnerable people, untargeted scraping of facial images to build recognition databases, and emotion recognition in workplaces and schools (with limited exceptions).

2. High Risk

High-risk AI systems are allowed but carry the heaviest obligations. They include AI used in recruitment and hiring, credit scoring, education and exam scoring, access to essential services, law enforcement, border control, and AI that acts as a safety component in regulated products such as medical devices and machinery.

3. Limited Risk (Transparency Duties)

These systems must be transparent with users. Chatbots must tell people they are talking to AI, and AI-generated or manipulated content such as deepfakes must be clearly labeled.

4. Minimal Risk

Most AI applications, such as spam filters, recommendation engines, and AI in video games, fall here. They face no specific obligations under the Act, though voluntary codes of conduct are encouraged.

What Changed With the Digital Omnibus on AI

The Digital Omnibus on AI is the first major amendment to the EU AI Act. It was proposed by the European Commission in November 2025, adopted by the European Parliament and the Council in June 2026, and published in the Official Journal in July 2026. It is now in force.

The main reason for the change was practical. The harmonised technical standards and conformity assessment setup needed for high-risk AI were not ready in time. Here are the key changes for businesses:

  • Standalone high-risk AI (Annex III) is delayed: obligations for systems used in recruitment, credit scoring, education, law enforcement, and border control now apply from 2 December 2027 instead of 2 August 2026.
  • Product-embedded high-risk AI (Annex I) is delayed further: AI used as a safety component in regulated products such as medical devices, machinery, toys, lifts, and vehicles now applies from 2 August 2028.
  • A narrower “safety component” test: AI that only assists users or optimizes performance does not automatically become high-risk if its failure poses no risk to health or safety.
  • Relief for smaller companies: the Act now includes formal definitions for SMEs and small mid-cap companies, with simplified technical documentation.
  • Bias detection: processing personal data is permitted where strictly necessary to detect and correct bias, with appropriate safeguards.

One point matters more than any headline about a “delay.” The Digital Omnibus is not a pause on AI regulation. Most transparency obligations were not delayed, the rules for general-purpose AI models still apply, and the EU AI Office has its full enforcement role. Companies that stop preparing now risk running out of time.

Updated EU AI Act Timeline (2025 to 2028)

DateWhat AppliesWho It Affects
2 February 2025Prohibited AI practices banned; AI literacy provisions beginAll providers and deployers
2 August 2025Obligations for general-purpose AI (GPAI) modelsFoundation and GPAI model providers
2 August 2026General application of the Act, transparency rules, and AI Office enforcementProviders and deployers of chatbots, generative AI, and other covered systems
2 December 2027High-risk AI obligations for Annex III systemsAI used in hiring, credit, education, law enforcement, border control
2 August 2028High-risk AI obligations for Annex I systemsAI embedded in regulated products such as medical devices and machinery
EU AI ACT Compliance timeline

Provider vs Deployer Obligations

Your obligations depend on your role. Providers carry most of the responsibility for high-risk systems, while deployers have narrower but still important duties. Note that a deployer can become a provider if it puts its own name on a system or substantially modifies it.

RequirementProviderDeployer
Risk management systemRequired across the full lifecycleUse the system according to instructions
Data governanceEnsure training data quality and relevanceEnsure input data is relevant
Technical documentationPrepare and maintainKeep records of use
Human oversightDesign the system to allow itAssign trained people to oversee it
LoggingBuild automatic logging into the systemRetain the logs
TransparencyProvide clear instructions for useInform affected people where required
Registration and conformityConformity assessment and EU database registrationRegister in specific public sector cases

Penalties for Non-Compliance

The EU AI Act has three tiers of fines. For large companies, the higher of the two amounts applies:

  • Prohibited AI practices: up to €35 million or 7% of global annual turnover.
  • Most other violations, including high-risk and transparency obligations: up to €15 million or 3% of global annual turnover.
  • Supplying incorrect or misleading information to authorities: up to €7.5 million or 1% of global annual turnover.

Example: a company with €200 million in annual turnover that breaches a high-risk obligation faces a maximum fine of €15 million, because that is higher than 3% of its turnover (€6 million). For SMEs and startups, the lower of the two amounts applies, so a qualifying SME with the same turnover would face a maximum of €6 million.

Where AI Agents and Generative AI Fit

The EU AI Act does not create a separate category for AI agents. Instead, an agent is classified by what it does. This makes agentic AI compliance a question of use case, not technology.

  • Customer-facing AI agents and chatbots usually fall under limited risk. They must tell users they are interacting with AI.
  • Autonomous AI agents that make decisions in high-risk areas, such as screening job candidates or approving loans, are treated as high-risk systems. They need human oversight, logging, and documentation.
  • Generative AI tools that create text, images, audio, or video must label AI-generated content so it can be identified as such.
  • Companies building on GPAI models such as GPT, Claude, or Gemini rely on the model provider for model-level obligations, but remain responsible for how their own application is used.

Because agents act with more autonomy, they need clear limits on what they can do, audit trails of their actions, and a human checkpoint for high-impact decisions. Building these controls in from the start costs far less than retrofitting them later.

Also Read: GDPR Compliance Consulting for AI and Software Projects

EU AI Act Compliance Checklist

Use this checklist to move from awareness to action:

  1. Build an AI inventory. List every AI system and model you develop, buy, or use, including AI features inside third-party tools.
  2. Classify each system by risk. Map it to prohibited, high, limited, or minimal risk, and check whether it falls under Annex III or Annex I.
  3. Define your role. Confirm whether you are a provider, deployer, importer, or distributor for each system.
  4. Remove prohibited uses. Confirm that none of your systems use banned practices.
  5. Run a gap analysis. Compare your current processes with the obligations that apply to each system.
  6. Assign a compliance owner. Name a person or team responsible for AI governance across legal, product, and engineering.
  7. Prepare technical documentation. Record system purpose, training data, design choices, testing, and known limitations.
  8. Design human oversight. Add review steps and override controls, especially for AI agents making high-impact decisions.
  9. Add transparency notices. Disclose AI interactions and label AI-generated content.
  10. Review your vendors. Check that AI vendors and model providers can support your compliance needs.
  11. Train your teams. Make sure staff who build or use AI understand its risks and limits.
  12. Monitor and review. Set up logging, incident reporting, and regular reviews as systems and rules change.

How Xicom Helps You Build Compliant AI

Compliance is easier when it is part of how AI is built, not something added at the end. At Xicom, we follow a compliance-by-design approach across our AI development services and AI agent development work. That includes risk classification at the planning stage, clear technical documentation, audit logging, and human-in-the-loop workflows for high-impact decisions.

Xicom holds ISO/IEC 42001, ISO/IEC 27001, and SOC 2 Type II certifications. Our AI governance consulting team helps businesses assess existing AI systems, close compliance gaps, and build governance frameworks that keep pace with EU AI Act updates.

Conclusion

The Digital Omnibus gave businesses more time for high-risk AI, but it did not reduce what is required. Transparency rules and general-purpose AI obligations already apply, and the high-risk deadlines in December 2027 and August 2028 are closer than they seem. EU AI Act compliance works best when you start with an inventory, classify your systems, and build oversight into your products now. If you need help assessing your AI systems or building new ones that meet the rules.

For the official text, see the EU AI Act on EUR-Lex and the European Commission’s AI Act page.

FAQs

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

Yes. The EU AI Act applies based on where AI is used, not where a company is based. If your AI system is placed on the EU market or its output is used in the EU, you may need to comply, even if your company is in India, the US, or elsewhere.

When do high-risk AI rules apply after the Digital Omnibus?

Under the Digital Omnibus, obligations for standalone high-risk AI systems listed in Annex III apply from 2 December 2027. High-risk AI embedded in regulated products under Annex I, such as medical devices and machinery, applies from 2 August 2028.

Are AI chatbots high-risk under the EU AI Act?

Most AI chatbots are limited risk, not high-risk. They must clearly tell users they are interacting with AI. A chatbot can become high-risk if it is used for high-impact decisions, such as screening job applicants or assessing creditworthiness.

What are the penalties for EU AI Act non-compliance?

Fines reach up to €35 million or 7% of global turnover for prohibited practices, up to €15 million or 3% for most other violations, and up to €7.5 million or 1% for supplying incorrect information. SMEs pay the lower of the two amounts.

What is the difference between a provider and a deployer?

A provider develops an AI system and places it on the market under its own name. A deployer uses an AI system in its professional work. Providers carry most obligations for high-risk systems, while deployers must use systems correctly, ensure human oversight, and keep logs.

Do AI agents need to comply with the EU AI Act?

Yes. The Act has no separate category for AI agents, so each agent is classified by its use case. A customer support agent is usually limited risk, while an agent making decisions in hiring or lending is high-risk and needs oversight, logging, and documentation.

Rahul Mahajan
The Author

Rahul Mahajan

Founder and CEO · Xicom
With over two decades of experience leading technology and business strategy, Rahul Mahajan has shaped the AI and digital transformation direction of enterprises across industries including Healthcare, Retail, FinTech, and Education. Under his leadership as the Founder and CEO of Xicom, the company has scaled to a 350+ member team and delivered 1800+ projects for clients across 50+ countries.

Make your ideas turn into reality
With our AI & mobile app solutions

Discuss your project with our experts

9 + 7 =