Regulation (EU) 2024/1689
EU AI Act compliance audit: classifying an AI system by risk and the application timetable
The EU Artificial Intelligence Act entered into force on 1 August 2024 and applies in stages until 2 August 2027. It is not addressed to model providers alone: a deployer using an AI system in recruitment or credit assessment falls within Annex III. Sila classifies your system, identifies the stage that applies to it, and builds the corresponding compliance plan.
By Younès Lahrach, data protection specialist and external DPO, director of Sila International Advisory. First published on , last reviewed against the texts on .
Who is caught
Article 2 covers, among others, providers placing AI systems on the market or putting them into service in the Union, irrespective of where they are established, as well as deployers established in the Union. It also covers providers and deployers established in a third country where the output produced by the system is used in the Union.
What is classified is not the company but the system, assessed case by case. The Regulation sets up four distinct regimes of decreasing intensity.
Prohibited practices, Article 5
Prohibition of, among others, subliminal or manipulative techniques materially distorting behaviour, exploitation of vulnerabilities relating to age, disability or a social or economic situation, social scoring by or on behalf of public authorities, untargeted scraping of facial images to build facial recognition databases, and emotion recognition in the workplace and in education, outside medical and safety uses.
High-risk systems, Article 6 and Annexes I and III
Two routes to classification. Article 6(1) covers AI systems intended to be used as safety components of products covered by the harmonisation legislation in Annex I, or which are themselves such products subject to third-party conformity assessment. Article 6(2) refers to the eight areas of Annex III, including biometrics, critical infrastructure, education and vocational training, employment and workforce management, access to essential private and public services including creditworthiness assessment, law enforcement, migration and asylum, and the administration of justice.
General-purpose AI models, Chapter V
Obligations of technical documentation, information to downstream providers, a copyright policy, and publication of a summary of the content used for training (Article 53). Article 51(2) presumes systemic risk where the cumulative amount of computation used for training, measured in floating point operations, is greater than 10^25, which triggers the enhanced obligations of Article 55.
Transparency obligations, Article 50
Informing natural persons that they are interacting with an AI system, machine-readable marking of synthetic content, information where an emotion recognition or biometric categorisation system is used, and labelling of deep fake content.
Application timetable, Article 113
The Regulation entered into force on 1 August 2024. Article 113 sets four application dates, and those dates govern the sequencing of any compliance plan.
2 February 2025
Chapters I and II apply: general provisions, the prohibitions in Article 5, and the AI literacy obligation in Article 4, which requires providers and deployers to ensure a sufficient level of AI literacy among their staff.
2 August 2025
Chapter III Section 4 (notifying authorities and notified bodies), Chapter V (general-purpose AI models), Chapter VII (governance), Chapter XII (penalties) with the exception of Article 101, and Article 78 (confidentiality) apply. Providers of general-purpose AI models placed on the market before that date have until 2 August 2027 to comply, under Article 111(3).
2 August 2026
The Regulation applies as a whole, subject to the exceptions above and below. This stage covers in particular the high-risk systems falling under Annex III by virtue of Article 6(2), and the transparency obligations of Article 50.
2 August 2027
Article 6(1) and the corresponding obligations apply, namely high-risk systems that are safety components of products covered by the harmonisation legislation in Annex I.
Penalties and competent authority
Article 99 sets the ceilings applicable to providers and deployers. Each ceiling takes the higher of a fixed sum and a percentage of total worldwide annual turnover for the preceding financial year. Article 99(6) reverses that rule for SMEs, including start-ups: the ceiling is then the lower of the two.
Article 99(3) provides for administrative fines of up to EUR 35 000 000 or, if the offender is an undertaking, up to 7 % of its total worldwide annual turnover, whichever is higher, for non-compliance with the prohibition of the AI practices in Article 5. Article 99(4) provides for up to EUR 15 000 000 or 3 % for non-compliance with the obligations it lists, and Article 99(5) for up to EUR 7 500 000 or 1 % for the supply of incorrect, incomplete or misleading information to notified bodies or national competent authorities.
How Sila engages
The engagement begins with an inventory of the AI systems actually in use, which rarely matches the list held by the IT function, and with classification of each of them under Article 6: prohibited practice, high risk, general-purpose model, transparency obligation only, or outside the Regulation.
Classification determines the applicable stage and therefore the deadline that concerns you, as well as the allocation of obligations between you and your providers. Article 25 reverses that allocation in three cases, and a deployer that changes the intended purpose of a system may find itself bearing the provider's obligations.
The compliance plan then follows the sequence set out below, and the evidence file is built to be produced to the market surveillance authority.
The method, in four steps
01Gap assessment
Establishing whether, and on what basis, the instrument applies to the activity as it is actually carried out, then comparing what the instrument requires against what already exists inside the organisation. The assessment ends with a ranked list of gaps, not an overall score. An organisation that falls outside the scope learns so at this stage, and the engagement stops there.
02Mapping
Recording what the measures will apply to: processing activities and data flows, systems, service providers, contracts in force, roles and responsibilities. The map is the working document for everything that follows, and it is the document produced to a supervisory authority in the event of an inspection.
03Technical and organisational measures
Implementing the measures the mapping has shown to be necessary: internal procedures, policies, contractual clauses to be renegotiated with providers, governance decisions, and training for the people concerned. Measures that sit inside the information system are specified by Sila and implemented by your own teams or your IT provider.
04Evidence file
Assembling the file that can be produced to a supervisory authority: records, impact assessments, dated and versioned policies, decision trails, proof of training. Compliance is not demonstrated by the absence of an incident but by the production of this file, which means it has to be kept current once the engagement ends.
A tailored quotation within 48 hours
The scope of an engagement cannot be guessed at: it depends on whether and how the instrument applies to you, on the number of processing activities, systems and contracts involved, and on what already exists inside your organisation. No amount is therefore published on this site.
The process starts with a scoping discussion, at no charge and without commitment, to define what the engagement will cover. A tailored quotation follows within 48 hours, setting out the scope, the deliverables and the amount.
Frequently asked questions
Does the AI Act apply to a company that only uses AI tools?
Yes, as a deployer. Article 3(4) of Regulation (EU) 2024/1689 defines the deployer as any natural or legal person using an AI system under its own authority, except where used in the course of a personal non-professional activity. Article 26 sets out obligations specific to deployers of high-risk systems, and Article 4 imposes an AI literacy obligation on all providers and deployers regardless of the risk level.
What makes an AI system high risk?
Article 6 of Regulation (EU) 2024/1689 provides two routes. Under Article 6(1), the system is a safety component of a product covered by the Union harmonisation legislation listed in Annex I, or is itself such a product, and that product is subject to third-party conformity assessment. Under Article 6(2), the system falls within one of the eight areas listed in Annex III. Article 6(3) allows the classification to be set aside in narrowly defined circumstances, subject to documenting the assessment and registering the system.
When does the AI Act apply in full?
Article 113 of Regulation (EU) 2024/1689 provides for application in stages. Chapters I and II applied from 2 February 2025, Chapter V and the governance and penalty provisions from 2 August 2025, the Regulation as a whole from 2 August 2026, and Article 6(1) together with its corresponding obligations from 2 August 2027.
Can a deployer become a provider?
Yes, in the three situations set out in Article 25(1) of Regulation (EU) 2024/1689: putting its name or trademark on a high-risk system already placed on the market, making a substantial modification to such a system, or modifying the intended purpose of a system, including a general-purpose AI system, so that it becomes high risk. The deployer then bears the provider obligations of Chapter III Section 2.
What this regime triggers elsewhere
A system falling under Annex III almost always processes personal data, so the impact assessment under Article 35 of Regulation (EU) 2016/679 is then due, and Article 26(9) of Regulation (EU) 2024/1689 expressly provides that information produced under the AI Act may contribute to it.
Carrying out the associated GDPR impact assessmentAn AI system used as a safety component in the management of critical infrastructure falls under Annex III point 2 of Regulation (EU) 2024/1689, and the entity operating it most often falls within one of the sectors in Annex I of Directive (EU) 2022/2555.
Checking whether NIS2 appliesThe officer has no competence under the AI Act, which provides for no equivalent function, but is in practice the person who holds the record of processing activities on which the classification of the system is built.
Appointing a data protection officerReference texts
The information on this page is general in nature and does not constitute legal advice on a specific situation. Sila is not a law firm and does not provide representation before the courts.