EU AI Act · Cyprus and the European Union

EU AI Act Compliance in Cyprus: Five Service Lines

Philarist is a management and AI governance consultancy that runs the consulting work for EU AI Act compliance in Cyprus, in five service lines: an AI register, use-case triage, fundamental rights impact assessment, vendor terms and an AI policy. Legal opinions are given only by independent practising advocates. For high-risk systems listed in Annex III, such as AI credit scoring, the deployer obligations apply from 2 December 2027.

Key facts

Six facts to know before you start

The EU AI Act, Regulation (EU) 2024/1689, puts duties on those who use AI systems as well as those who build them. Using one under your own authority makes you a deployer (Art. 3(4)).

Philarist runs the consulting work. Legal opinions come from an independent practising advocate, your own or one we introduce you to, who advises you directly.

Key facts, with the article for each

  • AI systems intended to evaluate the creditworthiness of natural persons or establish their credit score are high-risk, except systems used to detect financial fraud (Annex III, point 5(b)).
  • Deployers of high-risk systems must assign human oversight to people with the necessary competence, training and authority (Art. 26(2)).
  • Deployers of high-risk systems keep the logs the system generates automatically, to the extent they control them, for at least six months unless other Union or national law provides otherwise (Art. 26(6)).
  • Before deploying a credit-scoring system, its deployer must assess the impact on fundamental rights, and may cross-refer to its GDPR data protection impact assessment (Art. 27(1) and (4), as replaced in 2026).
  • The Annex III high-risk obligations apply from 2 December 2027 (Art. 113, third paragraph, point (c), as amended by Regulation (EU) 2026/1744).
  • The Advocates Law, Cap. 2, counts giving an opinion on any legal matter submitted to the advocate as practising as an advocate (s.2(1)(vi)), and only a registered advocate holding an annual licence may practise (s.11(1)).

The five service lines

Five service lines, one workflow

1. AI register

List every AI system, its role and risk.

Philarist builds the inventory and runs the workshops. An advocate writes the legal classification memo, for example “high-risk under Annex III”.

Read more about the AI register

2. AI use-case triage

Approve, condition or refuse each AI use.

Philarist designs the process and reviews each use technically. An advocate records the legal decision and its conditions.

Read more about AI use-case triage

3. Fundamental rights impact assessment

Assess the effect on people’s fundamental rights.

Philarist runs the process, gathers the facts and supplies the template and technical risk analysis. An advocate gives the legal fundamental-rights assessment and sign-off.

Read more about the impact assessment

4. AI vendor terms

Review and negotiate the AI supplier’s contract.

Philarist carries out technical due diligence on the supplier: data, logs and model changes. An advocate reviews and negotiates the contract and its clauses.

Read more about AI vendor terms

5. AI policy

The client’s internal rules for using AI.

Philarist prepares the policy as a governance document, trains staff and runs the annual review. An advocate reviews it against the AI Act, GDPR and sector rules.

Read more about the AI policy

Who does what

Who does what: the full split

Service linePhilarist (consulting)Independent practising advocate
1. RegisterInventory of AI systems, workshops, the registerLegal classification memo (for example, “high-risk under Annex III”)
2. TriageProcess design, technical review of each useLegal decision record and its conditions
3. Impact assessmentRunning the process, gathering facts, the template, technical risk analysisLegal fundamental-rights assessment and sign-off
4. Vendor termsTechnical due diligence on the AI supplier (data, logs, model changes)Contract review, negotiation and clauses
5. PolicyAI policy as a governance document, training, annual reviewLegal review against the AI Act, GDPR and sector rules

The Advocates Law, under legal sources below, is the reason for the split.

Worked example

A Cyprus lender that scores applicants with AI

The lender in this example is generic, not a real company. It uses a vendor’s credit-scoring model to score individual applicants.

That use is on the Annex III high-risk list, point 5(b), which excepts fraud detection. Using the model under its own authority makes the lender a deployer (Art. 3(4)).

Diagram headed ’A Cyprus lender scores creditworthiness with AI’: three individual applicant profile cards lead through an arrow to an ’AI credit score’ gauge, then through a second arrow to a gold box reading ’High-risk, Annex III, point 5(b)’.
A Cyprus lender that scores individuals’ creditworthiness with AI uses a high-risk system (Annex III, point 5(b)).

Three duties from 2 December 2027

As deployer of a high-risk system, the lender must, among other duties, assess the impact on fundamental rights (Art. 27), assign human oversight (Art. 26(2)) and keep logs for at least six months (Art. 26(6)).

Slide headed ’From 2 December 2027’ showing three boxes for what the lender, as deployer of a high-risk system, must do: a fundamental rights impact assessment (Art. 27), human oversight (Art. 26) and keeping logs at least six months (Art. 26(6)); a line below says the law firm advises and the lender is the deployer.
The lender’s three duties from 2 December 2027. On the slide, “the law firm” means the lawyers who advise the lender. On this page that is an independent practising advocate; Philarist is not a law firm.

The five steps in the lender’s case

1

Register

The lender lists every AI system it uses, with its role and risk, starting with the scoring model; fraud detection is excepted from point 5(b).

2

Triage

Credit scoring is approved on conditions: a person reviews every refusal, and applicant files never go into personal AI accounts. Art. 26(2) is the duty behind the reviewers; any right to human intervention in a solely automated refusal comes from GDPR Art. 22, where it applies, not the AI Act.

3

Impact assessment

Who is affected, what can go wrong and how a refused applicant can complain (Art. 27(1)(c), (d) and (f)). It may cross-refer to the lender’s GDPR data protection impact assessment (Art. 27(4), as replaced by Regulation (EU) 2026/1744).

Three stacked slides for the first three service lines in a Cyprus lender’s case. Register: list every AI system, its role and risk; credit scoring is high-risk under Annex III 5(b) and fraud detection is excepted. Triage: approve, condition or refuse each AI use; credit scoring is conditional on human review and applicant files in personal AI accounts are never allowed. Impact assessment: assess the effect on people’s fundamental rights; who is affected, what can go wrong and how a refused applicant can complain.
Steps 1 to 3 for the same lender.
4

Vendor terms

Is the lender’s data used for training? Who is liable for a wrong score? What happens when the model changes? The lender keeps logs only to the extent they are under its control, and as a financial institution keeps them as part of the documentation it keeps under Union financial services law (Art. 26(6)), so access to them is a contract question.

5

Policy

Trained reviewers for credit decisions, and an update when the model or the law changes. The reviewers’ training and authority come from Art. 26(2). Keeping the policy current is good governance rather than a separate AI Act duty.

Two stacked slides for the last two service lines in a Cyprus lender’s case. Vendor terms: review and negotiate the AI supplier’s contract; is the lender’s data used for training, and liability for a wrong score and model changes. Policy: the client’s internal rules for using AI; trained reviewers for credit decisions, and a policy updated when a model or the law changes.
Steps 4 and 5 for the same lender. On every step Philarist runs the consulting work and an advocate gives the legal opinion. On the slide, “the firm” means the advisers, not necessarily Philarist.

Why now

Why start before 2 December 2027

For a deployer of an Annex III high-risk system, such as AI credit scoring, the date to plan against for the high-risk obligations is 2 December 2027.

Other AI Act provisions already apply. Chapters I and II apply from 2 February 2025, except the new Art. 5 prohibitions in Art. 5(1), points (ba) and (bb), and Art. 5(1a) and (1b), which apply from 2 December 2026 (Art. 113, third paragraph, point (a), as amended). Art. 4, on AI literacy, sits in Chapter I; Regulation (EU) 2026/1744 replaced it.

  • 2 August 2026: the AI Act’s general date of application.
  • 2 December 2027: the Annex III high-risk obligations, including the deployer duties in Articles 26 and 27 (Art. 113, third paragraph, point (c)(i)). Annex I systems follow on 2 August 2028.

The register comes first because the other four lines depend on it. An impact assessment is due before a high-risk system is deployed (Art. 27(1)).

Systems already in service are treated differently. Under Art. 111(2), as amended, the Regulation applies to operators of high-risk systems placed on the market or put into service before the date of application of Chapter III only if, from that date, those systems are subject to significant changes in their designs. Providers and deployers of high-risk systems intended for use by public authorities must in any case comply by 2 August 2030. Whether a vendor’s model update meets that test is a question for an advocate.

As at 12 October 2026, the Commission’s draft guidelines on high-risk classification are not final, and the AI Office’s impact-assessment template (Art. 27(5)) is not yet published. Cyprus’s implementing legislation has so far been published only as draft bills: see the legal sources below.

How engagement works

How an engagement runs, from first call to handover

1

First call and scope

We talk through the AI systems you use or plan to use and which of the five lines apply, then agree in writing what is in scope and where an advocate’s opinion is needed. If a line does not apply, we say so.

2

Your advocate

Use your own advocate, or we introduce you to one. The advocate advises you directly and is responsible for the legal opinion. An introduction is not a joint offer.

3

Consulting work

Philarist runs the workshops, builds the register, designs the triage process, gathers the facts and does technical due diligence on suppliers.

4

Legal opinion

The advocate gives the classification, the decision records, the fundamental-rights assessment and sign-off, and the contract review and negotiation, and reviews the policy.

5

Handover and upkeep

You receive the register, templates, process and policy, and training. Recurring lines continue as ongoing work.

Fees: shapes, not prices

Philarist’s consulting fees are typically a fixed fee for a one-off project such as a register or an impact assessment, and a retainer for recurring work such as triage and policy upkeep.

This page states no prices. Scope and fee are agreed before work starts.

Tools

Tools: Claude for Legal as an example

Anthropic publishes Claude for Legal, an open-source collection of 13 plugins under the Apache-2.0 licence. Its ai-governance-legal plugin (version 1.0.2, checked on 12 October 2026) has skills for an AI inventory, use-case triage, impact assessments, vendor AI review and a starter AI policy.

We show it as an example of a tool, not as an endorsement. Its own guidance says its default frameworks are often US-centric, so each output has to be adapted to Cyprus and EU law, and its README says every output is a draft for attorney review. In Cyprus that means a draft for an advocate’s review. The code is on GitHub.

Claude and Anthropic are trademarks of Anthropic PBC, shown for identification only, with no endorsement or partnership implied.

Videos

Watch: the EU AI Act as a service line

EU AI Act for Law Firms: A New Service Line, the Lender Example
EU AI Act for Law Firms: The 5 Changes That Matter Most
AI Register for Law Firms: What It Is and How to Build One

Swipe for more videos →

Both videos were made with AI, using an avatar and voice clone of Philippos Aristotelous; the content and views are his own.

FAQ

Questions about the EU AI Act and these five service lines

Is AI credit scoring high-risk under the EU AI Act?

Yes: Annex III, point 5(b), lists AI systems intended to evaluate the creditworthiness of natural persons, or establish their credit score, as high-risk, except systems used to detect financial fraud. A Cyprus lender using such a system is normally its deployer. Whether a particular system is covered is a classification for an advocate.

When do the high-risk obligations apply?

The Annex III high-risk obligations apply from 2 December 2027 (Art. 113, as amended by Regulation (EU) 2026/1744). For high-risk systems already on the market or in service before the date of application, the Regulation applies to their operators only if the systems undergo significant changes in their design from that date; systems intended for public authorities must comply by 2 August 2030 (Art. 111(2), as amended).

What must a lender that uses AI credit scoring do under the AI Act?

As deployer of a high-risk system, the lender must assign human oversight to people with the necessary competence, training and authority (Art. 26(2)), keep the logs under its control for at least six months (Art. 26(6)) and assess the impact on fundamental rights before deploying it (Art. 27(1)). These apply from 2 December 2027.

What is a fundamental rights impact assessment?

A fundamental rights impact assessment is the assessment of a high-risk AI system’s impact on fundamental rights that Article 27 requires before deployment. It applies to certain public bodies and private entities providing public services, and to deployers of credit-scoring or life and health insurance systems (Annex III, points 5(b) and (c)). It covers the people affected, the risks of harm and complaint arrangements.

Does the EU AI Act give a refused loan applicant a right to human review?

No: the AI Act does not give that right. GDPR Art. 22 governs decisions based solely on automated processing with legal or similarly significant effects, such as an automated credit refusal; where the decision is necessary for a contract or based on explicit consent, the controller must provide at least the right to obtain human intervention (Art. 22(3)). The AI Act requires human oversight by people with the necessary competence, training and authority (Art. 26(2)). Making a person review every refusal is a condition set in triage.

Can a consultant give legal opinions on the AI Act in Cyprus?

No: a legal opinion on the AI Act in Cyprus has to come from an advocate, not from a consultancy. The Advocates Law, Cap. 2, counts giving an opinion on any legal matter submitted to the advocate as practising as an advocate (s.2(1)(vi)), and only a registered advocate holding an annual licence may practise (s.11(1)). Practising without being registered, or without a current annual licence, is an offence (s.11(3)). A consultancy can run the process.

Do we need our own lawyer?

Yes, for a legal opinion on the AI Act: it has to come from an independent practising advocate, your own or one Philarist introduces you to. The advocate advises you directly and is responsible for the opinion. Philarist’s AI register and process design do not replace it.

Can Claude for Legal do the EU AI Act work?

Not on its own: Claude for Legal produces drafts, and an advocate has to review them. Its ai-governance-legal plugin has skills for an AI inventory, use-case triage, AI impact assessments, vendor AI review and a starter AI policy. Its own guidance says its default frameworks are often US-centric, its README says every output is a draft for attorney review, and its impact-assessment skill treats a fundamental rights impact assessment as a separate deliverable. In Cyprus, an advocate has to review and adapt any draft.

Related pages: the AI register, triage, the impact assessment, vendor terms, the AI policy, legaltech & AI implementation for law firms, Claude for law firms, Contact and My Story.

Philippos also designs and delivers an HRDA-approved programme for Cyprus law firm leaders, the Law Firm Digital Transformation Framework.

Written by Philippos Aristotelous, founder of Philarist. Last reviewed: 12 October 2026.

This page is general information about the EU AI Act and related law. It does not constitute legal advice and should not be relied on as such. Philarist is a management and AI governance consultancy, not a law firm, and does not give legal advice or legal opinions. Legal opinions, legal review and contract negotiation are provided by independent practising advocates.

Next step

Planning for the EU AI Act? Book a call, or email info@philarist.com.

Book a call