EU AI Act fines: Penalty tiers and enforcement explained
Understand the EU AI Act fines and enforcement timeline. A breakdown of the penalty tiers, the SME proportionality rule, and the enforcing authorities.
In this article8

The penalty framework and enforcement timeline
Regulation (EU) 2024/1689 establishes a harmonised AI framework across all 27 EU Member States, and Chapter XII carries its sanctions. The penalties are not flat-rate: they scale with the severity of the infringement, the role the entity plays in the value chain, and the size of the undertaking.
The enforcement regime takes effect in staggered phases under Article 113[1]. The Article 5 prohibitions became binding on 2 February 2025. Chapter XII, which contains Article 99, applies from 2 August 2025; the Commission's own power to fine general-purpose AI model providers under Article 101 applies only from 2 August 2026. The obligations the fines attach to phase in separately, and the Digital Omnibus on AI (Regulation (EU) 2026/1744, in force 27 July 2026)[3] moved the high-risk dates to 2 December 2027 for stand-alone Annex III systems and to 2 August 2028 for high-risk AI embedded in Annex I products.[2]
| Penalty tier | Statutory reference | Maximum fine (standard enterprise) | Fines available from |
|---|---|---|---|
| Tier 1: Prohibited practices | Article 99(3) | Up to 35,000,000 EUR or 7% of worldwide turnover | 2 August 2025 (prohibition binding since 2 February 2025) |
| Tier 2: High-risk non-compliance | Article 99(4) | Up to 15,000,000 EUR or 3% of worldwide turnover | 2 December 2027 (Annex III) / 2 August 2028 (Annex I) |
| Tier 2: GPAI model violations | Article 101 | Up to 15,000,000 EUR or 3% of worldwide turnover | 2 August 2026 |
| Tier 3: Misleading information | Article 99(5) | Up to 7,500,000 EUR or 1% of worldwide turnover | 2 August 2025 |
The three thresholds are not alternative readings of one maximum: each attaches to a different category of infringement - 35,000,000 EUR or 7% for prohibited practices, 15,000,000 EUR or 3% for high-risk and GPAI obligations, 7,500,000 EUR or 1% for misleading information[4].[6] In every tier the reference base is the same: total worldwide annual turnover for the preceding financial year, not EU revenue.
Tier 1: Fines for prohibited AI practices
Tier 1 is the highest ceiling in the Act. Under Article 99(3), breaching the Article 5 prohibitions draws fines of up to 35,000,000 EUR or, for an undertaking, 7% of total worldwide annual turnover for the preceding financial year, whichever is higher[4].
Article 5 practices are absolute bans, not risk-managed categories. Unlike high-risk systems they cannot be regularised through conformity assessment, contractual safeguards or post-market monitoring.
- Cognitive manipulation: subliminal or deceptive techniques that distort behaviour and cause significant harm.
- Exploitation of vulnerabilities of age, disability or socio-economic situation.
- Social scoring leading to unjustified detrimental treatment.
- Individual criminal-offence prediction based solely on profiling or personality traits.
- Untargeted scraping of internet or CCTV imagery to build facial recognition databases.
- Emotion recognition in workplaces and education, outside medical or safety uses.
- Biometric categorisation to deduce protected characteristics.
- Real-time remote biometric identification in public spaces for law enforcement, outside exhaustive exceptions.
The eight original prohibitions have bound operators since February 2025 and have carried Article 99 fines since 2 August 2025, so national authorities can already open proceedings under this tier. The AI Omnibus added a ninth prohibition, on AI that generates non-consensual sexually explicit imagery or child sexual abuse material, which takes effect in December 2026.[2]
Tier 2: High-risk system compliance and GPAI models
Article 99(4) is where most enforcement will actually land. It reaches non-compliance by operators or notified bodies with any obligation other than the Article 5 prohibitions, and sets fines of up to 15,000,000 EUR or 3% of an undertaking's worldwide annual turnover, whichever of the two is higher[4].
The statutory scope of Article 99(4) encompasses failures across the entire AI supply chain, including the core AI Act provider definition and deployer duties. It covers:
- Article 16 - providers: risk management, technical documentation, data governance, record-keeping, human oversight.
- Article 22 - authorised representatives: a non-EU provider must mandate an EU representative.
- Articles 23 and 24 - importers and distributors: verifying CE marking, certificates and documentation.
- Article 26 - deployers: operating outside the instructions for use, not monitoring for serious incidents, or skipping a fundamental rights impact assessment where one is required.
- Articles 31, 33(1), (3), (4) and 34 - notified bodies: requirements and operational obligations in conformity assessment.
- Article 50 - transparency: disclosure when a system interacts with people or generates synthetic content.
Article 101 sets the same ceiling - 15,000,000 EUR or 3% of annual total worldwide turnover, whichever is higher - for providers of general-purpose AI models[5]. The Commission fines directly where a provider intentionally or negligently infringes the Regulation, fails to supply a document or information requested under Article 91, denies access for a model evaluation under Article 92, or fails to comply with a measure requested under Article 93.
| Dimension | Article 99(4) High-risk systems | Article 101 GPAI models |
|---|---|---|
| Target entities | Providers, deployers, importers, distributors, notified bodies | GPAI model providers, including systemic risk |
| Maximum fine | 15,000,000 EUR or 3% worldwide turnover (higher) | 15,000,000 EUR or 3% worldwide turnover (higher) |
| Primary enforcement body | National market surveillance authorities | European Commission (European AI Office) |
| Core compliance focus | Risk management, data governance, CE marking, FRIAs | Training disclosures, copyright policies, systemic evaluations |
Tier 3: Supplying misleading information
Tier 3 is the lowest ceiling and covers obstruction of the authorities. Under Article 99(5), the supply of incorrect, incomplete, or misleading information to notified bodies or national competent authorities in reply to a request draws fines of up to 7,500,000 EUR or, for an undertaking, 1% of total worldwide annual turnover for the preceding financial year, whichever is higher[4].[6]
It applies whether or not the system itself is compliant: misstated training-data provenance, omitted test results or an incomplete technical file are each enough on their own.
Article 99(7) requires all relevant circumstances of the case to be taken into account when setting the amount[4]. Among them:
- Nature, gravity and duration of the infringement, the number of people affected and the damage done.
- Whether the infringement was intentional or negligent.
- Action taken by the operator to mitigate the harm suffered by affected persons.
- The degree of cooperation with the national competent authorities.
- Fines already imposed on the same operator for the same infringement, or by other authorities under other Union or national law for the same conduct.
The SME, start-up and small mid-cap rule
The Act contains one explicit exception to the ceiling formula. Under Article 99(6), in the case of SMEs, including start-ups, each fine referred to in Article 99 shall be up to the percentages or the fixed euro amounts of paragraphs 3, 4 and 5, whichever of the two is lower[4].
The direction is what matters. For a large undertaking the formula picks the higher figure, so a multi-billion-euro group cannot treat the fixed cap as a cost of doing business. For an SME it picks the lower one.
For a standard undertaking every ceiling in Article 99 applies at whichever of the two figures is higher[4]. The Digital Omnibus on AI extended the same lower-of ceiling to small mid-caps, through a new Article 99(6a), for fines under Article 99(4) and (5).[3]
SME status follows Commission Recommendation 2003/361: staff below 250, with turnover of €50 million or less or a balance sheet total of €43 million or less[7]. Furthermore, under Article 99(1), the penalties laid down by Member States must take into account the interests of SMEs, including start-ups, and their economic viability[4].
Enforcement authorities and sanctioning powers
Enforcement runs on two tracks: Union-level bodies and national market surveillance authorities. Which one has competence depends on the system or model in question.
- European AI Office: sits inside the Commission and exercises its direct fining powers over providers of general-purpose AI models under Article 101.[5] From 2 August 2026 the Digital Omnibus also gave it exclusive competence over AI systems built on a GPAI model by the same undertaking, and over systems inside a designated very large online platform.[2] It has its own investigation and inspection powers and can impose periodic penalty payments of up to 5% of average daily worldwide turnover.[3]
- National competent authorities: under Article 70 each Member State designates at least one notifying authority and one market surveillance authority. They impose all three Article 99 tiers on operators active in their territory.
- European Data Protection Supervisor (EDPS): under Article 100 the EDPS may fine Union institutions, bodies, offices and agencies up to EUR 1 500 000 for breaching the Article 5 prohibitions and up to EUR 750 000 for other infringements[8].[10] Under Article 74(9) it is their market surveillance authority, except for the Court of Justice acting judicially[8].[9]
Market surveillance authorities have broad investigative powers. Article 74(12) gives them full access to documentation and to the training, validation and testing data sets, limited to what the task requires. Source code is narrower: Article 74(13) allows access only on a reasoned request, and only once testing based on the data and documentation has been exhausted or proved insufficient.[9] Where a system stays non-compliant, the national procedure in Article 79 leads to withdrawal or recall from the market. Those evaluations also touch cross-border data sovereignty requirements and general AI infrastructure governance standards.
Article 99(10) subjects these powers to procedural safeguards, including effective judicial remedies and due process. Article 101(2) requires the Commission to put its preliminary findings to the provider and hear it before fining, and Article 101(5) gives the Court of Justice unlimited jurisdiction to cancel, reduce or increase the fine[5].
Frequently asked questions
When do the EU AI Act fines become enforceable?
The Article 5 prohibitions have bound operators since February 2025, and Article 99 penalties have applied since 2 August 2025. The Commission's power to fine GPAI model providers under Article 101 applies from 2 August 2026. High-risk obligations were deferred by the 2026 AI Omnibus to 2 December 2027 (Annex III) and 2 August 2028 (Annex I).
What is the maximum fine under the EU AI Act?
35 million EUR or 7% of total worldwide annual turnover for the preceding financial year, whichever is higher. Article 99(3) reserves this ceiling for the prohibited practices in Article 5.
How does the AI Act calculate fines for SMEs?
Article 99(6) reverses the formula: an SME or start-up is fined at whichever is lower of the euro amount and the turnover percentage. The 2026 AI Omnibus extended that relief to small mid-caps for the Article 99(4) and (5) tiers.
What are the fines for general-purpose AI model providers?
Up to 15 million EUR or 3% of annual total worldwide turnover, whichever is higher, imposed by the Commission under Article 101 for intentional or negligent infringements.
What happens if a company provides incorrect information to an authority?
Article 99(5) sets a separate ceiling of 7.5 million EUR or 1% of total worldwide annual turnover, whichever is higher.
Sources
- Article 113: Entry into Force and Application artificialintelligenceact.eu
- AI Act digital-strategy.ec.europa.eu
- Digital Omnibus on AI: Regulation (EU) 2026/1744 Is Published in the Official Journal nicfab.eu
- Article 99: Penalties ai-act-service-desk.ec.europa.eu
- Article 101: Fines for providers of general-purpose AI models ai-act-service-desk.ec.europa.eu
- Article 99: Penalties artificialintelligenceact.eu
- SME definition single-market-economy.ec.europa.eu
- Article 100: Administrative fines on Union institutions, bodies, offices and agencies ai-act-service-desk.ec.europa.eu
- Article 74: Market surveillance and control of AI systems in the Union market ai-act-service-desk.ec.europa.eu
- Article 100: Administrative Fines on Union Institutions, Bodies, Offices and Agencies artificialintelligenceact.eu
Rankings and scores on Optinest cannot be bought — no commercial relationship changes how a provider is assessed. How we score
Related reading
All Regulation →EU AI Act fines reach 7% of global turnover
An analytical guide to EU AI Act fines and enforcement. Learn how penalty tiers reach up to 35 million EUR and which authorities oversee market compliance.

AI Act enforcement: who fines whom, how much and from when
Who enforces the EU AI Act, the four fine tiers from EUR 7.5 million to EUR 35 million, and which dates apply after Regulation (EU) 2026/1744.

The AI Act's deadlines just moved
Track the EU AI Act compliance deadlines and obligations. Understand the 2026 enforcement timeline and the 2027 extensions for high-risk AI systems.
