EU AI Act compliance, one AI system at a time
The Act applies to each AI system separately. In Modulos you set the role and risk class of a system. The platform then selects the requirements that apply, assigns the controls and keeps the evidence current.
2 Aug 2026in effect
The Act applies in general, with transparency duties and fines for general-purpose AI models
2 Dec 2026in 10 days
New prohibitions apply and the marking grace period for older systems ends
2 Dec 2027in 12 months
High-risk duties for stand-alone systems
2 Aug 2028in 20 months
High-risk duties for AI inside regulated products
Check what the Act asks of one AI system
The Act runs four separate tests on every system: prohibited practices, high-risk use, transparency and general-purpose AI. The duties from each test add up. Pick the example closest to one of yours.
A model that scores loan applicants in the EU
High-risk
Creditworthiness assessment is on the Annex III list, point 5(b).
Duties apply by 2 December 2027
- Run risk management for the life of the systemArt. 9
- Govern training and test data, and check it for biasArt. 10
- Keep a technical file that stays currentArt. 11
- Log what the system doesArt. 12
- Give people the means to oversee itArt. 14
- Operate a quality management systemArt. 17
- Pass a conformity assessment before market entryArt. 43
- Register the system in the EU databaseArt. 49
Provider duties. A deployer carries a shorter list under Article 26. Roles in the docs
Risk assessment for life and health underwriting
High-risk
Risk assessment and pricing for life and health insurance is on the Annex III list, point 5(c).
Duties apply by 2 December 2027
- Run risk management for the life of the systemArt. 9
- Govern training and test data, and check it for biasArt. 10
- Keep a technical file that stays currentArt. 11
- Log what the system doesArt. 12
- Give people the means to oversee itArt. 14
- Operate a quality management systemArt. 17
- Pass a conformity assessment before market entryArt. 43
- Register the system in the EU databaseArt. 49
Provider duties. A deployer carries a shorter list under Article 26. Roles in the docs
A customer chatbot built on a licensed model
Transparency duties only
It talks to customers, so Article 50 applies. It is not on the high-risk list.
Duties apply since 2 August 2026
- Tell people they are talking to an AI systemArt. 50
- Give the staff who run it AI literacyArt. 4
A fine-tuned image model in a creative suite
Transparency duties, and a prohibition to rule out
It generates synthetic images, and the new ban on non-consensual intimate content applies unless adequate safeguards prevent it.
Duties apply by 2 December 2026
- Put technical safeguards against prohibited content in placeArt. 5
- Mark generated images in a machine-readable wayArt. 50
- Keep provenance in the generation pipeline
An in-house general-purpose model behind your own assistant
General-purpose AI duties
You provide a general-purpose model, so the Chapter V duties apply, and the assistant on top of it has transparency duties.
Fines apply since 2 August 2026
- Keep technical documentation for the modelArt. 53
- Publish a copyright policy
- Publish a summary of the training data
- Mark generated contentArt. 50
- Answer to the AI Office
The compliance file goes out of date when an AI system changes
The Act asks for an inventory that stays current, a classification you can defend for each system, controls with owners and evidence with dates.
A spreadsheet is correct on the day you complete it, but a retrained model or a new use case makes it out of date.
Turn the high-risk articles into controls with owners
The EU AI Act framework breaks each article into requirements and maps them to controls. Each control has an owner, a status and its evidence. One job assesses all of them on the schedule you choose, and the platform records the date of each run.
Job
EU AI Act · Claims triage assistant
Control assessment · Weekly · Monday 09:00
- Art. 9Risk management system14 controlsAssessed
- Art. 10Data and data governance11 controlsAssessed
- Art. 11Technical documentation6 controlsAssessed
- Art. 12Record-keeping5 controlsRunning
- Art. 14Human oversight8 controlsQueued
- Art. 15Accuracy, robustness and cybersecurity12 controlsQueued
31 of 56 control steps done in this run, for one AI system. Illustrative data.
Is the technical documentation for this AI system complete, current and kept under version control?
Evidence read
- model-card-v3.mdCited for scope and intended use
- system-architecture.pdfCited for components and data flow
- training-data-sheet.xlsxProposed mapping to this control
Control Assessment Agent
78% readiness
- The model card has no version history or change log.model-card-v3.md, section 1
- The data sheet does not state a retention period for training data.training-data-sheet.xlsx, tab Sources
Add a dated change log to the model card and record the retention period in the data sheet.
The score and findings are for your team to review. The control status stays yours to set.
Keep the technical file current between audits
The Control Assessment Agent reads the evidence attached to a control, scores its readiness and lists each gap with a citation. The score is the agent's proposal, but the control status stays yours to set.
See which controls reopen when a system changes
Runtime tests read live values from your monitoring. When a test fails, each control linked to it moves back to not executed, the reason goes into the activity log, and the control owner is notified.
- Mon 31 AugExecutedEvidence attached and locked
- Tue 22 SepTest failedDrift above the threshold on the new channel
- Tue 22 SepNot executedControl owner and test assignee notified
- Thu 24 SepEvidence refreshedRetrained model, new baseline
- Fri 25 SepExecutedSet by the control owner after the 06:00 pass
Comments and LogsTest Drift metric failed. Status updated from executed to not executed in control MCF-67.
Reopened in EU AI Act, NIST AI RMF. Illustrative data.
Fines scale with worldwide turnover
The higher of the two figures applies, but SMEs and start-ups take the lower one.
- €35M or 7%
- Prohibited practices
- €15M or 3%
- Most other duties
- €7.5M or 1%
- Incorrect information to an authority
Most of the Act's controls already serve another framework
102 of the 132 controls in the EU AI Act framework are also mapped to at least one other framework in the library. If you already follow NIST AI RMF, part of this work is done.
Controls shared with EU AI Act, of 132 · library 1.1.0
- EN 1828636
The quality management system standard for the EU AI Act. EN 18286 in the docs
- CRA17
ISO/IEC 42001 governs the organization, but the Act governs each AI system. The two share no system-level controls, so you need both.
Contributions to AI standards and regulation
The harmonized standards set out how a provider shows conformity with the Act. Two Modulos experts work on them.
Kevin SchawinskiCo-founder and CEO
Andrea BassoChairman of the Advisory Board
- JTC 21
- Experts in the committee that drafts the harmonized standardsKevin Schawinski and Andrea Basso serve as experts in CEN-CENELEC JTC 21 Working Group 3, Engineering aspects, and Working Group 4, Foundational and societal aspects.
- 2024
- ISO/IEC 42001 product conformity certificationThe Modulos AI Governance Platform holds the first product conformity certification for an AI governance platform against ISO/IEC 42001:2023, issued by the accredited body CertX (certificate 213-001/24).The ISO/IEC 42001 page
- 2023–2025
- Spain's AI regulatory sandboxA Modulos team member served as an External Expert Advisor to Sandbox IA, Spain's pilot of the first EU AI Act regulatory sandbox, which concluded with the 16 AESIA compliance guides published in December 2025.
- Since 2024
- EU AI Pact signatoryModulos is a signatory of the EU AI Pact, the European Commission's voluntary pledge to apply AI Act principles ahead of full applicability.
Questions about the EU AI Act
What is the EU AI Act?
The EU AI Act (Regulation 2024/1689) is the European Union’s comprehensive law on artificial intelligence, in force since 1 August 2024. It is a product safety regulation: it bans some AI practices outright, requires conformity assessments and CE marking for high-risk AI systems, imposes transparency obligations, and sets model-level obligations for general-purpose AI providers.
When did the EU AI Act come into force?
The EU AI Act came into force on 1 August 2024. Obligations apply on staggered dates: Article 5 prohibitions and Article 4 AI literacy from 2 February 2025; Chapter V general-purpose AI model-provider obligations from 2 August 2025; general application on 2 August 2026. Since the July 2026 amendment (Regulation (EU) 2026/1744), Annex III high-risk obligations apply from 2 December 2027 and Annex I product-integrated high-risk obligations from 2 August 2028.
Is the EU AI Act in force right now?
Yes. The AI Act has been legally in force since 1 August 2024. Already applicable: Article 5 prohibitions and Article 4 AI literacy since February 2025; Chapter V general-purpose AI model-provider obligations since August 2025; and general application, Article 50 transparency and Commission GPAI fine powers included, since 2 August 2026. Annex III high-risk obligations apply from 2 December 2027 and Annex I product-integrated high-risk obligations from 2 August 2028, dates set by the July 2026 amendment (Regulation (EU) 2026/1744).
Is the EU AI Act a regulation or a directive?
A regulation. It applies directly across all 27 EU member states without national transposition, on the staged application dates in Article 113. This is the same legal instrument as the GDPR, and member states cannot soften or vary its core obligations, though they still legislate nationally for authorities, penalties, and enforcement arrangements.
What does the EU AI Act regulate?
It bans a small set of AI practices outright; requires high-risk AI systems to undergo conformity assessment and CE marking before being placed on the market; imposes transparency obligations on AI systems that interact with people or generate synthetic content; and sets model-level obligations for general-purpose AI providers, including documentation and copyright-related disclosures.
What practices are prohibited under the EU AI Act?
Article 5 prohibits social scoring of natural persons that meets the conditions in Article 5(1)(c); manipulative or exploitative AI that causes harm; real-time remote biometric identification in public spaces by law enforcement (with narrow exceptions); biometric categorization based on sensitive characteristics; emotion recognition in workplaces and education; predictive policing based solely on profiling; and untargeted facial-image scraping for facial recognition databases. From 2 December 2026, Article 5(1)(ba) and (bb) add AI generating non-consensual intimate imagery and child sexual abuse material, with provider and deployer conditions set in Article 5(1a).
What are high-risk AI systems under the EU AI Act?
Two routes under Article 6. Annex I: AI systems that are safety components of products covered by listed Union harmonization legislation (medical devices, machinery, vehicles, radio equipment, toys, and others), where Article 6(1)(b) requires the product to undergo a third-party conformity assessment. Annex III: stand-alone AI systems in eight domains, namely biometrics, critical infrastructure, education, employment, essential public and private services, law enforcement, migration and border control, and the administration of justice and democratic processes.
Who does the EU AI Act apply to?
Article 2 sets the scope: providers placing AI systems on the EU market (regardless of where the provider is established); deployers established or located in the Union; importers and distributors of AI systems in the EU; product manufacturers placing AI systems on the EU market; authorized representatives of non-EU providers; and providers and deployers in third countries when the AI system outputs are used in the Union.
Does the EU AI Act apply to UK companies?
Yes, if the UK company places an AI system on the EU market or its AI system outputs are used in the EU. Article 2 hooks providers through placing on the EU market, and both providers and deployers where the system’s output is used in the Union, wherever the company is headquartered. The UK’s post-Brexit status does not exempt UK companies from EU AI Act obligations when EU users or markets are involved.
Does the EU AI Act apply to US companies?
Yes. A US company selling an AI system into the EU market, or whose AI system outputs are used by EU customers, is in scope. A US provider placing a high-risk AI system on the EU market must appoint an authorized representative established in the EU under Article 22. Non-EU providers of general-purpose AI models have a separate authorized-representative rule under Article 54.
How do I comply with the EU AI Act?
Inventory your AI systems; classify each against the four independent gates (prohibited, high-risk, transparency, GPAI); confirm your role under Articles 3 and 25 (provider, deployer, importer, distributor). If you are a provider of a high-risk AI system, the Article 16 stack applies: run a conformity assessment under Article 43, build the Article 11 technical file, draw up the EU Declaration of Conformity, affix the CE mark under Article 48, and operate a quality management system under Article 17 (EN 18286 is the harmonized standard for it); register the system under Article 49 where required (mainly Annex III, including systems self-exempted from high-risk classification under Article 6(3), which stay registrable). Deployers operate under Article 26 instead, with a narrower set of duties unless Article 25 reclassifies them as providers.
What are the penalties for EU AI Act non-compliance?
Up to €35 million or 7% of global annual turnover for prohibited-practice violations, whichever is higher. Up to €15 million or 3% for most other obligations. Up to €7.5 million or 1% for supplying incorrect, incomplete, or misleading information to authorities. SMEs and start-ups take the lower of the fixed amount and the percentage under Article 99(6), and small mid-caps get the same lower-of rule for the two non-prohibition tiers under Article 99(6a). National market surveillance authorities enforce most penalties; Article 101 fines on GPAI model providers are imposed by the European Commission directly.
How does the EU AI Act relate to ISO/IEC 42001?
ISO/IEC 42001 is the international management system standard for AI. Article 17 of the AI Act requires providers of high-risk AI systems to operate a quality management system. The harmonized standard for it is EN 18286: since 14 October 2026, conformity with EN 18286 gives a presumption of conformity with Article 17(1). ISO/IEC 42001 is not a substitute, but many AI Act programs operate inside an ISO/IEC 42001 management system.
How does the EU AI Act relate to NIST AI RMF?
NIST AI RMF is voluntary US guidance for AI risk management. The EU AI Act is binding EU regulation with legal obligations and penalties. Implementing NIST AI RMF builds the risk-management practices the Act expects, particularly Article 9 risk management for high-risk providers, but it does not substitute for the Act’s specific compliance and conformity-assessment requirements.
Has the EU AI Act been delayed?
Yes, for high-risk systems, and the new dates are fixed. The July 2026 amendment (Regulation (EU) 2026/1744, in force since 27 July 2026) shifted Annex III stand-alone high-risk obligations from 2 August 2026 to 2 December 2027 and Annex I product-integrated high-risk obligations from 2 August 2027 to 2 August 2028. It also added a new Article 5 prohibition on non-consensual intimate content and CSAM (compliance by 2 December 2026), set the Article 50 watermarking grace deadline at 2 December 2026 for systems placed on the market before 2 August 2026, and moved the Article 57(1) date for national regulatory sandboxes to 2 August 2027. The prohibitions timetable, the GPAI duties, and the 2 August 2026 general-application date stayed in place.
How does Modulos help with EU AI Act compliance?
Modulos automates the compliance workflow for high-risk AI systems: AI system inventory, classification across the four gates, Article 9 risk management, Article 10 data governance, Article 11 technical documentation, Article 13 transparency, Article 17 quality management, and ongoing monitoring. The platform holds CertX product conformity certificate 213-001/24 against ISO/IEC 42001:2023, and the framework library includes EN 18286, the harmonized standard for the Article 17 quality management system.
Bring one AI system to the demo
Pick a system you are not sure about. We set its role and risk class with you and show the requirements, controls and evidence the Act asks for.
GDPR, NIS2, DORA and Cyber Resilience Act
- GDPR Personal data in training, inputs and outputs, and automated decisions about people.
- NIS2 Cybersecurity measures and incident reporting for the infrastructure your AI runs on.
- DORA ICT risk, incident reporting and third-party oversight for financial entities.
- Cyber Resilience Act Security duties for software and hardware products sold in the EU.
In your sectorFinancial servicesTransportationUtilitiesMobilityTelecommunicationsDefense and national securityExpert panel on the Digital Omnibus