Three numbers determine your exposure under the EU AI Act:
- the fine tier,
- your company’s annual turnover,
- and whether you qualify for the SME cap.
Everything else is arithmetic.
Article 99 of Regulation (EU) 2024/1689 sets three maximum fine tiers. The highest, €35 million or 7% of total worldwide annual turnover, applies to AI systems that are outright prohibited. For most companies using AI in recruitment, credit assessment, insurance pricing, or customer-facing applications, the relevant ceiling is €15 million or 3%. The percentage is calculated on total worldwide revenue for the preceding financial year, not on EU-only revenue or the revenue attributable to the AI system itself.
There is one variable most fine estimates ignore. Article 99(6), which reverses the calculation for companies with fewer than 250 employees and turnover below €50 million. For qualifying companies, the rule is the lower of the EUR ceiling or the percentage figure, not the higher. That distinction can reduce a theoretical maximum by 90% or more. The calculator below applies this rule automatically, based on your answers.
This tool is designed for compliance officers, general counsel, and COOs who need a number before a board meeting, not a legal opinion. It uses the fine tiers from Article 99 and Article 101, applies the SME definition from Commission Recommendation 2003/361/EC, and flags the enforcement dates from Article 113. It does not replace legal advice.
How this calculator works
The calculator maps your AI system’s primary function to the correct fine tier under Article 99, or to the separate GPAI regime under Article 101 for providers of general-purpose AI models. It applies the SME inverse cap from Article 99(6) automatically: companies with fewer than 250 employees and turnover below €50 million get the lower of the EUR ceiling or the percentage, where large companies face the higher of the two. The tool does not cover Article 99(5) fines for supplying false information to regulators, Article 100 fines for EU institutions, individual fine calibration factors under Article 99(7), or member state legislation that may add criminal penalties on top of the administrative regime.
EU AI Act Fines: Common Questions
What is the maximum fine under the EU AI Act?
The maximum fine is €35 million or 7% of total worldwide annual turnover for the preceding financial year, whichever is higher. This applies to violations of Article 5, which covers prohibited AI practices: social scoring, real-time biometric identification in public spaces, emotion recognition at the workplace, and systems designed to manipulate behaviour below the threshold of conscious awareness. For companies with fewer than 250 employees and turnover below €50 million, Article 99(6) applies the lower of the two figures rather than the higher.
When do EU AI Act fines start applying?
Article 99 fines became applicable on 2 August 2025 under Article 113(b). The Article 5 prohibitions on which Tier 1 fines are based took effect from 2 February 2025. Fines under Article 101 for general-purpose AI model providers apply from 2 August 2026. Annex III high-risk obligations, which trigger Tier 2 fines for recruitment AI, credit scoring, and similar systems, apply from 2 August 2026 under the general application date in Article 113. A provisional political agreement between the European Council and Parliament on 7 May 2026 would defer the Annex III date to 2 December 2027, but this is not yet binding law.
Does the SME fine reduction apply to my company?
If your company has fewer than 250 employees and annual turnover of €50 million or less, or a balance sheet total below €43 million, it qualifies as an SME under Commission Recommendation 2003/361/EC. Article 99(6) then replaces the standard “higher of EUR ceiling or percentage” rule with the lower of the two. A company with 80 employees and €8 million in turnover faces a maximum Tier 2 fine of €240,000, not €15 million. This reduction applies only to Article 99 fines. It does not apply to Article 101 fines for general-purpose AI model providers, where the higher of the two always applies regardless of size.
What types of AI system trigger EU AI Act fines?
Three categories trigger administrative fines under Article 99. Prohibited practices under Article 5, in force from 2 February 2025, are the highest-risk: deploying any of the eight listed prohibited AI systems exposes you to Tier 1 fines. High-risk AI systems listed in Annex III, including recruitment tools, credit scoring, insurance pricing, healthcare triage, and education admissions systems, trigger Tier 2 fines from 2 August 2026 once Annex III obligations apply. AI systems with transparency obligations under Article 50, including chatbots and systems generating synthetic content presented to humans as real, fall under the same Tier 2 ceiling.
Who enforces EU AI Act fines?
For most violations under Article 99, fines are imposed by the national market surveillance authority in the Member State where the company operates or places its AI system on the market. Member States designate their own authorities; enforcement capacity and priorities vary. Providers of general-purpose AI models face a different enforcement path: Article 101 fines are imposed exclusively by the European Commission via the EU AI Office, not by national authorities. The Court of Justice of the EU has unlimited jurisdiction to review, reduce, or increase Article 101 fines under Article 101(5).
Does the EU AI Act apply to companies outside the EU?
Yes. The EU AI Act applies to any provider, deployer, importer, or distributor whose AI system is placed on the EU market or used within the EU, regardless of where the company is established. A US company whose recruitment AI is used by an EU employer is in scope. The fine is calculated on total worldwide annual turnover, not on EU revenues or revenues attributable to the AI system. This is the same approach the EU uses for GDPR and competition fines.
How do EU AI Act fines compare to GDPR fines?
GDPR’s maximum is €20 million or 4% of worldwide annual turnover, whichever is higher. The EU AI Act’s top tier is €35 million or 7%, also whichever is higher. For prohibited AI practices, the AI Act ceiling is materially larger than GDPR’s. For high-risk obligations and transparency requirements, the €15 million or 3% ceiling sits below GDPR’s maximum percentage but above the EUR ceiling. Article 99(8) addresses overlap: where a violation also breaches other EU law, only the higher applicable fine is imposed for the same facts.
What happens if the Digital Omnibus on AI is formally adopted?
A provisional political agreement reached on 7 May 2026 would defer the Annex III standalone high-risk AI obligations from 2 August 2026 to 2 December 2027, and the Annex I product-embedded high-risk category from 2 August 2027 to 2 August 2028. It would also introduce a new “small mid-cap” company category extending some SME reliefs to slightly larger businesses. Until the agreement is formally adopted, expected by July 2026, the original Article 113 dates remain in force. If the deadline passes without adoption, the 2 August 2026 date applies as written.
Next step
If the calculator returned a significant fine exposure, the next step is understanding which specific obligations apply to your AI system. The AI Act Readiness Assessment covers the full set of obligations by system type, with a plain-language checklist for each.
EU compliance deadlines are moving targets in 2026. The fine amounts are fixed in the regulation. The dates on which they bite are still being negotiated. Subscribe to RegDossier for monthly updates when enforcement dates change, new guidance appears, or member states start issuing the first penalties.
Subscribe to RegDossier — EU regulatory updates in your inbox.
This tool provides general guidance based on Regulation (EU) 2024/1689 (EU AI Act). It does not constitute legal advice. EU regulations are subject to member state transposition and interpretation. Consult qualified professionals for your specific situation.