EU AI Act Compliance Guide for Malta and Cyprus (2026)
Many businesses heard that the AI Act had been delayed. That’s partially true. The Digital Omnibus on AI, in force since 27 July 2026, moved most high-risk duties to 2 December 2027. However, several important rules already apply. These include bans on certain AI practices, requirements to help employees understand AI, and rules about telling people when AI is being used.
EU AI Act compliance depends on how risky your use of AI is. The Act sorts AI into four tiers, from banned practices to minimal-risk tools, and each tier carries different duties. Most local businesses use AI as deployers, so the first job is a list of every AI system in use, its purpose and the data behind it. In this article, we will dive deeper into the tiers, deadlines and steps that companies can take now.
Quick Answer: What Does the EU AI Act Require From Businesses?
The EU AI Act, formally Regulation (EU) 2024/1689, is the European Union’s law on artificial intelligence. It entered into force on 1 August 2024 and applies in stages. The full text of the EU AI Act is published on EUR-Lex. As an EU regulation, it applies directly in Malta and Cyprus, with no separate national law needed for the duties to take effect.
What your business needs to do depends on two things: your role and the level of risk.
A provider builds an AI system or sells it under its own name. A deployer uses an AI system in its work.
For example, a Maltese bank using a supplier’s chatbot is a deployer.
Most businesses are deployers rather than providers. Their obligations are generally less extensive, although some uses of AI still carry important responsibilities.
As of November 2026, these are the duties that matter for most businesses:
- Banned practices. Since 2 February 2025, practices such as social scoring and emotion recognition in the workplace are prohibited.
- AI literacy. Since 2 February 2025, providers and deployers must take measures to build AI literacy among staff who use AI. In the Digital Omnibus on AI the rules were simplified in July 2026 and do not require businesses to guarantee a particular level of knowledge for every employee.
- Transparency. Since 2 August 2026, people must be told when they are talking to an AI system, and deepfakes and certain AI-generated content must be labelled.
- High-risk duties. These apply from 2 December 2027 for stand-alone systems, such as AI used in hiring or credit checks. AI built into regulated products follows from 2 August 2028.
Fines reach €35 million or 7% of worldwide annual turnover for banned practices, whichever is higher. Most other breaches carry up to €15 million or 3%. For small and medium-sized enterprises (SMEs), the lower of the two amounts applies.
Who Does the EU AI Act Apply To?
The Act covers any organisation that builds, imports, distributes or uses AI systems in the EU. It also reaches companies outside the EU when the output of their AI is used inside it.
For example, a Malta-licensed iGaming company serving customers in Germany or Finland may fall under the Act when using AI in its business. Where its servers are located does not, by itself, determine whether the Act applies.
The Act does not generally cover AI used purely for personal, non-professional purposes. It also includes exclusions for certain military, national security and research activities.
In Eunoia’s experience, businesses in Malta and Cyprus should start with practical questions about their data:
Where is our data stored and processed?
Who can access it?
Is our data being used to train AI models?
How can we check whether the AI’s answers are correct?
At Eunoia, we start by understanding what the AI system will do, what data it needs and what risks it could create. For example, using AI to forecast sales or extract information from documents is very different from using AI to assess employees or make decisions that affect people’s lives. These differences help determine which rules apply.
Who Enforces AI Regulation in Malta and Cyprus
Malta
AI regulation in Malta follows the EU AI Act, applied through two legal notices published in 2025. Legal Notice 226 of 2025 names the Malta Digital Innovation Authority (MDIA) as the lead market surveillance authority and single point of contact. The MDIA is also responsible for a national AI regulatory sandbox.
Legal Notice 227 of 2025 gives the Information and Data Protection Commissioner (IDPC) supervision of certain high-risk systems, such as those used in law enforcement and biometric identification. For financial services, the MDIA coordinates with the Malta Financial Services Authority. The MDIA’s artificial intelligence page sets out its role, and its enforcement powers became exercisable on 2 August 2026.
Cyprus
Cyprus designated its authorities by Council of Ministers decision on 22 January 2025. The Commissioner of Electronic Communications and Postal Regulation acts as the notifying authority, market surveillance authority and single point of contact.
The Commissioner for Personal Data Protection supervises specific high-risk categories, and the Deputy Ministry of Research, Innovation and Digital Policy coordinates national work. The Cyprus EU AI Act portal lists these roles.
The Four Risk Tiers: Which Applies to Your AI System?
A common misconception in Eunoia’s client conversations is that the Act treats every AI tool the same way, and that any AI project brings a heavy legal burden. That is not the case.
The table below sets out the four tiers, with the dates that apply after the Digital Omnibus.
| Risk tier | Definition | Examples | Key requirement | Compliance deadline |
|---|---|---|---|---|
| Unacceptable risk | AI practices that threaten people’s safety or fundamental rights | Social scoring; emotion recognition at work or in schools; manipulative techniques; untargeted scraping of facial images | Banned | 2 February 2025 (bans on AI-generated non-consensual intimate images and child abuse material added from 2 December 2026) |
| High risk | AI used for decisions with serious effects on people, or built into regulated products | Recruitment and staff evaluation; credit scoring of individuals; life and health insurance pricing; AI in medical devices | Risk management, data governance, logging, human oversight and registration; deployer duties under Article 26 | 2 December 2027 for stand-alone (Annex III) systems; 2 August 2028 for AI in regulated (Annex I) products |
| Limited (transparency) risk | AI that interacts with people or generates content | Customer service chatbots; AI-generated images, audio or video; deepfakes | Tell people they are dealing with AI; label or mark generated content | 2 August 2026; machine-readable marking for generative systems already on the market by 2 December 2026 |
| Minimal risk | Most other AI systems | Sales and demand forecasting; spam filters; stock planning; internal document data extraction (usually) | No new duties beyond AI literacy | AI literacy duty applies since 2 February 2025 |
Some AI uses can be classified as high-risk even when they seem like normal business activities.
For example, an insurance company using AI to assess risks or calculate prices for life or health insurance may be using a high-risk system. A similar tool used for motor insurance is not automatically included in that high-risk category.
Banks using AI to assess an individual’s creditworthiness also fall within the high-risk category, although AI used specifically for fraud detection is excluded from that particular classification.
For iGaming companies, AI used to group players for marketing purposes is not normally classified as high-risk under the AI Act. However, the General Data Protection Regulation (GDPR) still applies when personal data is involved.
Article 22 of the GDPR also restricts certain decisions made entirely by automated systems when they have legal or similarly significant effects on individuals.
In Eunoia’s work with player segmentation models, we recommend keeping marketing decisions separate from player protection rules. For example, customers who have self-excluded should not receive promotions simply because an AI model identifies them as likely to respond.
For profiling that creates higher privacy risks, businesses should also assess whether a data protection impact assessment is required.
What Malta and Cyprus Businesses Need to Do Now
Until July 2026, 2 August 2026 was the date most guides pointed to. The Digital Omnibus on AI moved the high-risk duties to December 2027. Enforcement of the bans and the AI literacy duty began on 2 February 2025, and the transparency rules on 2 August 2026.
Businesses should not wait until December 2027 to begin preparing.
At Eunoia, we recommend starting with the purpose of each AI system, the data it uses and the risks involved. Here are five practical steps.
- List every AI system in use. Include AI features inside existing software, such as Microsoft 365 Copilot, CRM tools and HR platforms. Record the purpose, the supplier, the data used and the internal owner.
- Classify each system. Decide whether your business is the provider or the deployer, and which risk tier applies. Check any system used for hiring, credit, insurance pricing or access to services against Annex III of the Act.
- Address the requirements that already apply. Stop any banned practices, add chatbot disclosures and label AI-generated content. The European Commission’s Article 50 questions and answers explain the transparency duties. Ask suppliers of generative AI tools how they will mark content by 2 December 2026, when the grace period for systems already on the market ends.
- Put role-based AI literacy measures in place for your employees. Record what training each group of staff received and when.
- Prepare high-risk systems for 2 December 2027. Deployers need human oversight by trained staff, logs kept for at least six months and notice to employees before high-risk AI is used at work. Banks and insurers using AI for credit scoring or life and health insurance pricing also need a fundamental rights impact assessment.
The Omnibus also simplified documentation for SMEs and small mid-cap companies. Malta has assigned the MDIA to run a national AI regulatory sandbox, where firms will be able to test systems under supervision.
These steps involve legal judgement as well as technical work. Eunoia’s role is on the technology side, and its AI integration services design and deploy AI with governance, security and validation built in. Where a project carries specific regulatory duties, Eunoia works alongside the client’s legal or compliance team, or brings in specialist external advice.
How AI Governance Connects to Your Data Strategy
Many of the EU AI Act’s requirements depend on how businesses manage their data. Providers must show that training data is relevant and documented. Deployers must keep logs, check input data and let people review outputs. A business that cannot trace where its data came from will struggle with each of these.
From Eunoia’s experience working with data projects, two governance controls deliver the most value early: centralised access control and data lineage. Together, they show who can access data, where it came from and which reports or models a change will affect. Tools such as Databricks Unity Catalog and Microsoft Purview record this information across the data estate. As a Microsoft Solutions Partner with a Databricks partnership, Eunoia sets up these controls on both platforms.
Explainability depends on the type of AI. In forecasting and predictive analytics work, Eunoia shows clients the data, the assumptions and how the model performs, alongside the forecast itself. In document intelligence, each extracted value comes with a confidence level and a highlighted area on the original page, so a reviewer can check it quickly. Eunoia’s q3 agent lets users ask questions of their Power BI semantic models, shows the query behind each answer and works within each user’s existing data permissions.
Insurance models follow the same principles in Eunoia’s approach. Each score comes with an explanation. A person can override the recommendation, and every override is logged for traceability. The model retrains when its performance falls below an agreed threshold.
These practices support the Act’s principles of transparency, traceability and human oversight. Each system also needs a named owner, a defined purpose and a clear point where a person reviews the output, however well it is built. Eunoia’s position is that AI governance belongs inside your data governance programme, with the same owners, catalogue and access rules. Running it as a separate legal exercise produces two sets of records that soon stop matching. When Eunoia’s data strategy consulting team plans a data platform, the AI inventory and the data catalogue are designed together.
Conclusion
The EU AI Act is now partly in force in Malta and Cyprus. Bans, the AI literacy duty and transparency duties apply today, and the MDIA, the IDPC and the Cypriot authorities can enforce them. High-risk duties apply from 2 December 2027 for stand-alone systems and from 2 August 2028 for AI in regulated products.
Most local businesses use AI tools developed by other companies, and many everyday uses of AI carry relatively few additional requirements under the Act.
However, every business should understand how it uses AI and which rules apply. A sensible starting point is to list your AI systems, understand the risks, assign responsibility, and make sure you can trace the data behind AI-generated results.
This will help you prepare for the EU AI Act while also making AI safer and more reliable for your business.
Build AI with clear governance from the start
Talk to Eunoia about building governed AI from the first release.
See governed reporting in a regulated firm
How Gordian Holdings automated daily regulatory reporting from one central warehouse
Check whether your data can support AI
A structured review of your data, access controls and documentation.
FAQ
Has the EU AI Act Been Delayed?
The Act has been partly delayed. The Digital Omnibus on AI, Regulation (EU) 2026/1744, entered into force on 27 July 2026. It moved high-risk duties for stand-alone systems to 2 December 2027 and for AI in regulated products to 2 August 2028. The bans, the AI literacy duty and the transparency duties already apply.