Ordit
Guide

The EU AI Act: a practical guide for regulated firms

After reading this you will be able to say whether the EU AI Act reaches your firm, which of your AI systems it touches, what you owe and by when, and how it fits with ISO/IEC 42001, GDPR and the rules your own regulator already applies.

Who this guide is for

  • Compliance officers and MLROs at regulated firms, whether supervised by the FSRA, the DFSA, the FCA, MAS or another authority, who have been asked by an EU client, counterparty or auditor whether the firm is "AI Act compliant".
  • Founders applying for a licence who plan to use AI in credit, insurance, onboarding or customer service.
  • Boards and senior managers who need a clear answer to "does this apply to us, and by when?".
  • Outsourced compliance providers and in-house counsel supporting firms with EU customers or EU group entities.

What is the EU AI Act?

The EU AI Act is Regulation (EU) 2024/1689, the European Union's comprehensive law on artificial intelligence. It regulates AI according to the risk of its specific use. Some practices are banned outright, some systems are classed as high-risk and carry detailed obligations, some must meet transparency duties, and most are left alone. The Act applies to the organisations that build or supply AI (providers) and the organisations that use it (deployers), wherever they are based, if the AI or its output reaches the EU.

The EU AI Act is Regulation (EU) 2024/1689, the European Union's comprehensive law on artificial intelligence.

The Act entered into force on 1 August 2024 and became generally applicable on 2 August 2026, with staggered dates for different chapters. It was substantively amended for the first time by Regulation (EU) 2026/1744, the Digital Omnibus on AI, in force from 27 July 2026. This guide reflects the amended text on 8 November 2026 and says so where a point is unsettled.

Does it apply to us?

The EU AI Act applies to a regulated firm outside the EU in two situations: when the firm supplies an AI system or general-purpose AI model into the EU market, and when the firm uses an AI system whose output is used in the EU. If neither applies, the Act does not regulate you, although your EU clients and counterparties may still ask you for AI governance evidence.

The scope is in Article 2(1). It covers providers placing AI systems or general-purpose AI models on the EU market, "irrespective of whether those providers are established or located within the Union or in a third country"; deployers established in the EU; and providers and deployers in a third country "where the output produced by the AI system is used in the Union". Importers, distributors, product manufacturers and authorised representatives are also covered.

Run each AI system you build, buy or use through these questions:

  1. Do you sell or license software with AI features to customers in the EU? You are likely a provider under Article 2(1)(a).
  2. Do you use AI in decisions about EU-resident customers, such as credit, insurance pricing or onboarding? Article 2(1)(c) may apply, because the output is used in the EU.
  3. Do you serve AI-generated content or an AI chatbot to users in the EU? Article 2(1)(c) may apply, and Article 50 transparency is the most likely obligation.
  4. Do you have an EU branch or subsidiary that uses AI? That entity is a deployer under Article 2(1)(b) in its own right.
  5. None of the above? The Act does not regulate you. Record the conclusion and revisit it when you add systems or markets.

Be careful with the second limb. Recital 22 frames it around output "intended to be used in the Union", commentators describe it as unsettled, and we found no Commission guidance for financial firms outside the EU. Record "may apply" and a documented position for each system.

The Act excludes military, defence and national security uses, scientific research, pre-market testing (except in real-world conditions), purely personal use, and free and open-source AI unless it is high-risk, prohibited or caught by Article 50. Non-EU providers of high-risk systems must also appoint an EU authorised representative under Article 22 (see the FAQ).

What does the EU AI Act require?

What are the four risk tiers?

The Act places every AI system in one of four tiers, and the obligations follow the tier. Unacceptable-risk practices are banned. High-risk systems, which could seriously affect health, safety or fundamental rights, carry detailed obligations for providers and deployers. Transparency-risk systems must tell people they are dealing with AI and mark synthetic content. Minimal-risk AI, which the Commission says covers most AI in use, has no specific rules. The Commission's own high-risk examples include credit scoring, CV-sorting software, exam scoring and remote biometric identification.

The Act places every AI system in one of four tiers, and the obligations follow the tier.

Which AI practices are prohibited?

Article 5 bans eight practices from 2 February 2025, and the Digital Omnibus on AI adds two more from 2 December 2026. The original eight are: harmful manipulative or deceptive techniques; exploiting vulnerabilities of age, disability or social or economic situation; social scoring; predicting criminal offending solely from profiling or personality traits; untargeted scraping of facial images for recognition databases; emotion inference in workplaces and education (medical and safety uses excepted); biometric categorisation inferring characteristics such as race, religion, trade union membership or sexual orientation; and real-time remote biometric identification in public spaces for law enforcement, with narrow exceptions.

The two added by the Omnibus are generating or manipulating non-consensual intimate imagery of identifiable people, and generating child sexual abuse material. For a financial firm the one most likely to surprise is emotion inference in the workplace, which rules out AI that reads staff mood or stress in call centres.

Which systems are high-risk, and which financial services uses are listed?

A system is high-risk under Article 6 by one of two routes: Annex I, for AI that is a safety component of a product already subject to third-party conformity assessment, and Annex III, for listed stand-alone use cases. Annex III is the route that matters for financial services.

Annex III, point 5, lists AI systems "intended to be used to evaluate the creditworthiness of natural persons or establish their credit score, with the exception of AI systems used for the purpose of detecting financial fraud", and AI systems "intended to be used for risk assessment and pricing in relation to natural persons in the case of life and health insurance". The other Annex III areas are biometrics, critical infrastructure, education, employment and worker management (including recruitment screening and performance monitoring), law enforcement, migration, and justice and democratic processes.

Two things follow. The fraud-detection carve-out is explicit and AML transaction monitoring is not in Annex III, so most compliance tooling is not high-risk, although Article 50, the general-purpose AI rules and a deployer's general duties can still apply. And HR uses are caught: if you use AI to screen CVs for roles in the EU, that is Annex III.

Article 6(3) lets a provider treat an Annex III system as not high-risk where it poses no significant risk, for example a narrow procedural or preparatory task, but never where it profiles natural persons; the assessment must be documented and the system registered.

What must a provider of a high-risk system do?

A provider of a high-risk system must meet the technical requirements in Section 2 of Chapter III, run a quality management system, complete a conformity assessment, register the system and keep demonstrating conformity. Article 16 lists the duties. The Section 2 requirements are a risk management system, data governance, technical documentation, record-keeping and logging, transparency and instructions for use, human oversight, and accuracy, robustness and cybersecurity. Article 17 adds the quality management system; the provider must also issue an EU declaration of conformity, affix CE marking, register in the EU database (Article 49) and take corrective action when needed.

What must a deployer of a high-risk system do?

A deployer of a high-risk system must use it as the provider instructs, assign human oversight to competent people, control the quality of input data, monitor the system, keep logs and tell people when it is used in decisions about them. Article 26 sets out the duties: follow the instructions for use; assign oversight to people with the competence, training and authority to intervene; ensure input data is relevant and representative where you control it; monitor for risks and report serious incidents; keep logs for at least six months; inform workers before workplace use; inform individuals that a high-risk system is used in decisions about them; use the provider's information for a GDPR Article 35 data protection impact assessment where one is required; and cooperate with authorities. Most regulated firms will be deployers. The list is shorter than the provider's, but it falls on the firm, not the vendor.

When is a fundamental rights impact assessment required?

A fundamental rights impact assessment is required under Article 27 before first use of a high-risk system by public bodies, by private entities providing public services, and by deployers of two listed uses: credit scoring of natural persons and risk assessment and pricing in life and health insurance. It must describe the processes, the period and frequency of use, the people affected, the specific risks of harm, the human oversight measures and the complaints arrangements, and Article 27(4) allows it to be combined with a data protection impact assessment. If you deploy credit scoring or life and health pricing to EU-resident customers, put this on the plan now.

What are the rules for general-purpose AI models?

Providers of general-purpose AI (GPAI) models have had obligations since 2 August 2025. Under Article 53 they must keep technical documentation, give downstream providers the information needed to understand the model's capabilities and limits, have a policy to comply with EU copyright law, and publish a sufficiently detailed summary of the training content. Open-source models are exempt from the first two duties unless they carry systemic risk. Article 51 presumes systemic risk when training compute exceeds 10^25 floating-point operations, with extra duties on evaluation, risk mitigation, incident reporting and cybersecurity. The Commission enforces these rules and, under Article 101, can fine GPAI providers up to EUR 15 million or 3% of worldwide turnover. Models on the market before 2 August 2025 have until 2 August 2027 to comply. Most regulated firms consume GPAI models through an API, so these duties sit with the supplier; know which models sit behind your tools and keep their documentation on file.

What are the Article 50 transparency duties?

Article 50 has applied since 2 August 2026 and is the obligation most firms will meet first. Providers and deployers must tell people they are interacting with an AI system unless it is obvious; mark synthetic audio, image, video and text output in a machine-readable, detectable way; disclose emotion recognition and biometric categorisation to those exposed to it; label deepfakes; and disclose AI-generated text on matters of public interest unless it has had human editorial review. Systems on the market before 2 August 2026 have until 2 December 2026 to meet the Article 50(2) marking duty. If your chatbot or marketing content reaches EU users, start here.

What did the Digital Omnibus on AI change?

The Digital Omnibus on AI, Regulation (EU) 2026/1744, moved the high-risk deadlines, added two prohibited practices, softened the AI literacy duty and extended relief to smaller firms. Obligations for Annex III high-risk systems now apply from 2 December 2027 instead of 2 August 2026, and for Annex I systems from 2 August 2028 instead of 2 August 2027. These are fixed dates, not conditional on harmonised standards as the Commission's original proposal had suggested.

The Commission tabled the proposal on 19 November 2025; provisional agreement followed on 6 May 2026, Parliament adopted the text on 16 June and the Council on 29 June, and it was signed on 8 July, published on 24 July and in force on 27 July 2026.

Other changes: the Article 4 AI literacy duty was softened from "ensure" to "take measures to support"; a new Article 4a allows special-category data processing to detect and correct bias, with safeguards; SMEs and small mid-caps get simplified technical documentation and the lower fine cap; the AI Office gains exclusive supervision of AI systems built on a provider's own GPAI model and of systems in very large online platforms; and the "safety component" definition was narrowed. Prohibitions, AI literacy, GPAI obligations and Article 50 kept their original dates.

What are the penalties?

Article 99 sets three tiers of fines, each the higher of a fixed sum or a share of total worldwide annual turnover. For prohibited practices: up to EUR 35 million or 7%. For most other obligations, including the high-risk duties of providers and deployers: up to EUR 15 million or 3%. For incorrect, incomplete or misleading information to authorities: up to EUR 7.5 million or 1%. SMEs and start-ups pay the lower figure, and the Omnibus extends that cap to small mid-caps for the 3% and 1% tiers. Member States enforce the Act for AI systems; the Commission enforces the GPAI rules.

Key dates and deadlines

Dates reflect the Act as amended by the Digital Omnibus on AI, as at 8 November 2026. The consolidated text of amended Article 113 was not fully available on EUR-Lex when we checked, so dates are taken from Recital 40 of the Omnibus and the AI Act Explorer's consolidated Article 113, which agree.

4 more dates in the table below.

Date What applies Status
1 August 2024 Regulation (EU) 2024/1689 enters into force Done
2 February 2025 Scope and definitions, prohibited practices (a) to (h), AI literacy (Article 4) Done
2 August 2025 Governance bodies, notified bodies, GPAI model obligations, most penalties Done
27 July 2026 Digital Omnibus on AI in force Done
2 August 2026 General application, including Article 50 transparency; national authorities begin enforcement Done
2 December 2026 New prohibitions (ba) and (bb); end of grace period for Article 50(2) marking on systems already on the market Pending
2 August 2027 GPAI models placed on the market before 2 August 2025 must comply; each Member State to have at least one AI regulatory sandbox Pending
2 December 2027 High-risk obligations for Annex III systems, including credit scoring and life and health insurance pricing (was 2 August 2026) Pending
2 August 2028 High-risk obligations for Annex I systems (was 2 August 2027) Pending
2 August 2030 Legacy high-risk systems used by public authorities must comply Pending

How it relates to ISO 42001 and GDPR

The short version: the AI Act is the legal obligation, ISO/IEC 42001 is the organisational governance layer and evidence base, EN 18286 is the emerging per-system conformity route for high-risk providers, and GDPR continues to apply to every piece of personal data an AI system touches.

Framework What it is How it relates to the AI Act What you should do
ISO/IEC 42001:2023 Certifiable management-system standard for AI, published December 2023, with clauses 4 to 10 and 38 Annex A reference controls Not a harmonised standard; certification gives no presumption of conformity. The European AI Office said in May 2024 that it was not fully aligned with the final Act Use it as the governance spine; map each AI Act article to a control and fill the gaps
EN 18286:2026 European standard for an AI quality management system for EU AI Act purposes, covering Article 17 The first harmonised standard, listed by Commission Implementing Decision (EU) 2026/2274 of 12 October 2026, published 14 October 2026. Presumption of conformity for Article 17 only Relevant if you provide a high-risk system. Its informative annex maps to ISO 42001, so 42001 work carries forward, but the mapping does not transfer the presumption
GDPR EU data protection law The AI Act is "without prejudice" to GDPR (Article 2(7)). Deployers use the provider's information for a GDPR Article 35 DPIA (Article 26(9)); the fundamental rights impact assessment can be combined with a DPIA (Article 27(4)); GDPR Article 22 limits solely automated decisions Treat the DPIA and the fundamental rights impact assessment as one exercise with two outputs. Check Article 22 whenever AI makes a decision with legal or similarly significant effect
Your local data protection law (for example the ADGM Data Protection Regulations 2021) The data protection law of your own jurisdiction Separate from GDPR. ADGM's regulations, for example, have their own automated decision-making, profiling and impact assessment provisions Applies regardless of the AI Act. Run the impact assessments together

No management-system certificate makes a high-risk system conform: conformity is assessed system by system against the Section 2 requirements, and in October 2026 only Article 17 had a harmonised standard. The Cloud Security Alliance lists five gaps a 42001 programme typically leaves against the Act: a per-system compliance strategy, change management for systems that keep learning, the Article 73 incident timelines, AI-specific supply-chain provisions and a structured fundamental rights assessment. If you hold ISO 42001, you are well placed, not finished.

What should a regulated firm outside the EU do about it?

Start from the honest position: outside the EU, the AI Act is one framework among several, and your own regulator's expectations on governance, outsourcing and technology risk apply whatever the EU does. Take the UAE as an example. It has no comprehensive AI law. Governance is policy-led (the National AI Strategy 2031, the 2024 UAE AI Charter and the 2022 AI Ethics Principles), and a federal AI and Data Authority was approved in June 2026. The Central Bank of the UAE issued a guidance note on responsible AI and machine learning for its licensed institutions in February 2026. ADGM has not introduced AI-specific regulations equivalent to the DIFC's Regulation 10, and at the time of writing we are not aware of an FSRA rule written specifically for AI; the hooks there are the ADGM Data Protection Regulations 2021, the FSRA's IT Risk Management Guidance (chapter 14 covers algorithm-driven solutions) and the GEN 3.5 cyber risk management rules. Whichever regulator you answer to, check for a dedicated AI rule or guidance, and expect your existing duties on governance, data protection and third-party risk to do most of the work.

So the EU AI Act regulates your firm only where you touch the EU market or EU people under Article 2(1)(a) or (c). Where you do, the obligations are real, Article 50 applies already, and the first hard deadline for the financial-services use cases is 2 December 2027. Either way, EU counterparties and group auditors are already asking for AI governance evidence, and the same work answers them and your own regulator's duties on governance and third-party risk. Build once, and let the AI Act be one of the frameworks your register serves.

A readiness checklist

  1. Inventory every AI system and model you build, buy or use, including AI features inside vendor software, and record where each system's output is used.
  2. Decide EU exposure for each system under Article 2: EU establishment, EU market, or output used in the EU. Write down the reasoning and use "may apply" where the position is unsettled.
  3. Screen every system against the Article 5 prohibitions, including the two that apply from 2 December 2026.
  4. Classify against Annex III, with particular attention to credit scoring, life and health insurance pricing and HR uses. Document any Article 6(3) derogation and remember that profiling is always high-risk.
  5. Map your role for each system: provider, deployer, importer or distributor. Article 25 can turn a deployer into a provider if you rebrand or substantially modify a vendor's system.
  6. For high-risk systems you deploy, assign named human oversight owners, keep logs for at least six months, prepare the individual notices, run the DPIA and, for credit and insurance, complete a fundamental rights impact assessment before first use.
  7. For high-risk systems you provide, build the Section 2 requirements and the quality management system, complete the conformity assessment, register in the EU database and appoint an EU authorised representative if you have no EU establishment.
  8. Meet Article 50 now: disclose AI interaction, mark synthetic content and label deepfakes. The marking grace period for existing systems ends on 2 December 2026.
  9. Put the whole thing inside a management system, ideally ISO/IEC 42001 integrated with your ISO 27001 ISMS and your regulator's risk management expectations, with board ownership and AI literacy measures for staff.
  10. Set calendar triggers for 2 December 2026, 2 August 2027, 2 December 2027 and 2 August 2028, each with a named owner and the evidence due.

Common mistakes

Assuming the Act stops at the EU border

Concluding that because the firm has no EU office, the Act is someone else's problem. Article 2(1)(a) catches any provider placing a system on the EU market wherever it is established, and Article 2(1)(c) can catch a deployer in Abu Dhabi, London or Singapore whose system's output is used in the EU. The right answer is a documented, per-system position.

Treating the 2027 delay as a reason to wait

The Omnibus moved the Annex III deadline to 2 December 2027. It did not move the prohibitions, the GPAI rules or Article 50, which apply now, and two new prohibitions plus the end of the marking grace period land on 2 December 2026. A credit scoring deployer also needs a fundamental rights impact assessment, human oversight and logging before first use, and thirteen months is not long for a bank.

Confusing ISO 42001 certification with conformity

ISO/IEC 42001 is valuable, and it is not a harmonised standard. Certification does not create a presumption of conformity, and even EN 18286 only covers the Article 17 quality management system. If a vendor tells you their certificate makes your use of their product "AI Act compliant", they are wrong twice: the certificate does not do that, and the deployer duties are yours regardless.

Missing the AI inside vendor software

Your inventory will be wrong if it only lists systems your own team built. Onboarding tools, CRM scoring, HR platforms, document review and customer service software increasingly contain AI features, some of which fall under Annex III. Ask every supplier what AI their product uses, where the output goes and which GPAI model sits behind it.

Forgetting the transparency duties that already apply

Article 50 has applied since 2 August 2026. A chatbot that does not say it is a chatbot, marketing imagery generated without machine-readable marking, or AI-drafted public content without editorial review are the easiest breaches to find and the easiest to fix.

Letting the deployer role drift into provider

Putting your own name on a vendor's AI system, substantially modifying it, or changing its intended purpose so that it becomes high-risk can make you the provider under Article 25, with the full Article 16 list attached. Check this before product and marketing teams rebrand a white-labelled tool.

Frequently asked questions

It can. Article 2(1)(a) applies to any provider placing an AI system or model on the EU market, wherever it is established. Article 2(1)(c) applies to providers and deployers in a third country where the system's output is used in the EU. If you serve EU customers, score EU residents or deploy AI through an EU group entity, assume the Act may apply and document a position for each system.

No, on the current text. Annex III does not list AML transaction monitoring, and the credit scoring entry expressly excludes systems used to detect financial fraud, so the high-risk obligations do not attach. The Article 50 transparency duties, the GPAI rules and general deployer duties can still apply, and the system should sit in your AI inventory with a documented classification.

For Annex III systems, which include credit scoring and life and health insurance pricing, the obligations apply from 2 December 2027. For Annex I systems, meaning AI that is a safety component of regulated products, from 2 August 2028. Both dates were set by the Digital Omnibus on AI in July 2026 and are fixed, not conditional on standards being ready.

No. ISO/IEC 42001 is not a harmonised standard and carries no presumption of conformity; the European AI Office said in May 2024 that it was not fully aligned with the final Act. Certification does show that your governance is structured and auditable, and EN 18286:2026 includes a mapping to it, so the work carries forward. Conformity for a high-risk system is still assessed system by system.

Not by using it. The Article 53 duties fall on the provider of the model. You become a provider of an AI system if you build a system on top of the model and place it on the EU market, and you may become a provider of a GPAI model if you modify the model itself. Keep your supplier's technical documentation and training summary on file.

Only if you are a provider of a high-risk AI system with no establishment in the EU. Article 22 requires you to appoint a representative in the EU by written mandate before placing the system on the market. The representative checks your documentation and conformity assessment, keeps records for ten years and is the contact point for authorities. Deployers do not need one.

If you deploy an Annex III point 5(b) or 5(c) system, that is credit scoring of natural persons or risk assessment and pricing in life and health insurance, Article 27 requires one before first use. Public bodies and private entities providing public services need one for any high-risk system. It can be combined with a data protection impact assessment, and most firms should run them together.

It depends on your jurisdiction, so check with your own regulator. In the UAE, for example, there is no comprehensive AI law as at November 2026. The approach is policy-led, with a federal AI and Data Authority approved in June 2026 and a Central Bank guidance note on responsible AI issued in February 2026. ADGM has no AI-specific regulation equivalent to the DIFC's Regulation 10; the relevant hooks are the Data Protection Regulations 2021, the FSRA IT Risk Management Guidance and GEN 3.5.

Key takeaways

  • The EU AI Act applies by reach, not by address. Work out exposure system by system under Article 2 and write it down.
  • Most compliance AI is not high-risk, but credit scoring, life and health insurance pricing and HR screening are, and profiling is always high-risk.
  • The dates that matter now are 2 December 2026 (new prohibitions, end of the marking grace period), 2 December 2027 (Annex III) and 2 August 2028 (Annex I). Article 50 already applies.
  • ISO/IEC 42001 is the right governance spine and gives no presumption of conformity. EN 18286:2026 is the first harmonised standard and covers Article 17 only.
  • Build one register of AI systems, obligations, controls and evidence that serves the AI Act, GDPR, local data protection law and your own regulator's duties together.

Sources

  1. European Commission, "AI Act" policy page, last updated 19 October 2026. https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai (checked 8 November 2026)
  2. EUR-Lex, Regulation (EU) 2026/1744 of 8 July 2026 (Digital Omnibus on AI), OJ L 2026/1744, 24 July 2026. https://eur-lex.europa.eu/eli/reg/2026/1744/oj (checked 8 November 2026)
  3. EUR-Lex, Commission Implementing Decision (EU) 2026/2274 of 12 October 2026 on harmonised standards (EN 18286:2026). https://eur-lex.europa.eu/eli/dec_impl/2026/2274/oj (checked 8 November 2026)
  4. AI Act Explorer, Article 2 (scope). https://artificialintelligenceact.eu/article/2/ (checked 8 November 2026)
  5. AI Act Explorer, Article 5 (prohibited practices, with Omnibus additions). https://artificialintelligenceact.eu/article/5/ (checked 8 November 2026)
  6. AI Act Explorer, Article 6 (classification) and Annex III. https://artificialintelligenceact.eu/article/6/ and https://artificialintelligenceact.eu/annex/3/ (checked 8 November 2026)
  7. AI Act Explorer, Articles 16, 22, 26 and 27 (provider, authorised representative, deployer and fundamental rights impact assessment). https://artificialintelligenceact.eu/article/16/ (checked 8 November 2026)
  8. AI Act Explorer, Articles 50, 51, 53, 99 and 101 (transparency, GPAI and penalties). https://artificialintelligenceact.eu/article/50/ (checked 8 November 2026)
  9. AI Act Explorer, Article 113 (consolidated application dates) and Digital Omnibus page. https://artificialintelligenceact.eu/article/113/ (checked 8 November 2026)
  10. Future of Privacy Forum, "The AI Act implementation timeline: what changes under the AI Omnibus", 28 July 2026, updated 5 August 2026. https://fpf.org/blog/the-ai-act-implementation-timeline-what-changes-under-the-ai-omnibus/ (checked 8 November 2026)
  11. Gibson Dunn, "EU AI Act Omnibus Agreement: Postponed High-Risk Deadlines and Other Key Changes", 27 May 2026. https://www.gibsondunn.com/eu-ai-act-omnibus-agreement-postponed-high-risk-deadlines-and-other-key-changes/ (checked 8 November 2026)
  12. Cloud Security Alliance, "EU AI Act Compliance: prEN 18286 and ISO 42001", 28 April 2026, and research note on the EN 18286 harmonised standard presumption, October 2026. https://labs.cloudsecurityalliance.org/research/csa-research-note-eu-ai-act-pren-18286-iso-42001-20260428/ (checked 8 November 2026)
  13. GDPR Article 22 text. https://gdpr-info.eu/art-22-gdpr/ (checked 8 November 2026)
  14. CMS, "Navigating AI in the UAE: a financial services perspective", 15 October 2026. https://cms.law/en/are/legal-updates/navigating-ai-in-the-uae-a-financial-services-perspective (checked 8 November 2026)
  15. ACA Group, "AI governance in UAE financial services is becoming a regulatory priority", 3 August 2026. https://www.acaglobal.com/industry-insights/ai-governance-in-uae-financial-services-is-becoming-a-regulatory-priority/ (checked 8 November 2026)
How Ordit helps

How Ordit helps

The EU AI Act asks for the same things your own regulator already asks for, applied to AI: know what you run, classify it, name who is accountable, keep evidence and prove it on request. Ordit treats the Act as one more framework in the register you already keep, not as a separate project.

From rulebook to evidenceHow this framework becomes a register you can prove, with people deciding at every step.
  1. 01Read the frameworkThe Act's articles and annexes, as amended by the Omnibus, are held as a set of obligations, each with its application date, so Article 50 shows as live today and Annex III shows as due on 2 December 2027.
  2. 02Map every requirementMap every requirement to one register of obligations, controls and evidence alongside your regulatory duties. Each AI system gets a record: role, EU exposure, risk tier, owner, assessment, controls and evidence of human oversight. Your policies are read against the duties, so the distance between what you say and what the Act requires is visible.
  3. 03Name an owner for each gapA missing fundamental rights impact assessment, an unmarked chatbot or a supplier with no model documentation becomes a task with a person and a date, not a line in a spreadsheet.
  4. 04Link controls and evidenceLink controls and evidence once so they serve every framework. The human oversight control that answers Article 26 is the same control your ISO 42001 auditor and your regulator's technology risk review will ask about. It is tested once, its evidence is fingerprinted and held in the immutable event chain, and it is shown through a scoped auditor or regulator view or exported in open formats.
  5. 05Review as it changesReview as the framework changes, with a named person accepting, amending or rejecting every change. When the Commission lists a new harmonised standard or a date moves again, horizon scanning proposes the change to the obligations and controls it affects, and nothing changes until someone in your firm decides.
Where it lives in OrditEach requirement on the left, the record or view that answers it on the right.
Article 2 scope and Article 6 classification
In OrditAI system record: role (provider or deployer), EU exposure, risk tier, Article 6(3) reasoning, owner
Article 5 prohibited practices
In OrditRegister of obligations: a prohibition screen on every AI system record, with evidence attached
Article 9 risk management system
In OrditRisk engine: AI risks linked to obligations, controls and key risk indicators, with scenario P50/P90/P95
Article 26 deployer duties (oversight, logs, notices)
In OrditAI system record: owner, assessment, controls, evidence of oversight, tested in the controls and testing programme
Article 27 fundamental rights impact assessment
In OrditAI system record assessment linked to the DPIA; evidence with fingerprints; sign-off in the immutable event chain
Article 50 transparency
In OrditControls and testing programme: disclosure and marking controls, tested, evidence attached
Articles 25 and 53: vendor AI and GPAI suppliers
In OrditSupplier and third-party record: model documentation, training summary, contract terms, incidents
Article 113 dates and Omnibus amendments
In OrditHorizon scanning and pack updates: each change proposed, a named person accepts, amends or rejects

Ordit proposes; your people decide, and the record shows who decided what and when.

This guide is for information only. It reflects the position on 8 November 2026 and is not legal or regulatory advice.

Book a discovery call
More guides

Plain answers to the frameworks you are asked about.

GuideADGM cyber risk management (FSRA GEN 3.5): a practical guide

FSRA GEN 3.5 cyber risk rules for ADGM firms, in force since 31 January 2026: governance, controls, suppliers, testing and 24-hour incident notification.

17 minute readRead the guide
GuideISO 42001: a practical guide for regulated firms

What ISO/IEC 42001 requires, whether a regulated firm needs it, how it fits with supervisory expectations and the EU AI Act, and how to get certified.

19 minute readRead the guide
GuideISO 27001: a practical guide for regulated firms

What ISO/IEC 27001:2022 requires, how a regulated firm scopes and certifies an ISMS, and how it maps to supervisory cyber rules and data protection law.

18 minute readRead the guide
GuideCyber Essentials: a practical guide for regulated firms

What Cyber Essentials covers, when a regulated firm in or outside the UK should get it, what it costs, why firms fail and what applies instead.

16 minute readRead the guide