In this article8 sections
  1. The Three Penalty Tiers
  2. The SME Rule: the Lower Figure Applies
  3. How the Amount Is Set Within the Ceiling
  4. Factors That Push Fines Up
  5. Factors That Push Fines Down
  6. What Enforcement Looks Like Today
  7. The One Number You Can Control
  8. References

A €35 million headline figure grabs attention. What it hides is that EU AI Act fine calculation is a structured process, and most of the inputs are things your business can influence before an investigation begins.

Understanding the mechanics is not just useful for lawyers. HR directors deciding whether to deploy an AI screening tool, compliance officers building an audit trail, and managing directors signing off on AI procurement all benefit from knowing how the enforcement math works.

The Three Penalty Tiers

The AI Act establishes three fine ceilings in Article 99.

Tier 1 applies to the most serious violations: placing on the market or using a prohibited AI practice listed in Article 5, such as social scoring, manipulative AI or emotion recognition in the workplace. The ceiling is €35 million or 7% of total worldwide annual turnover, whichever is higher.

Tier 2 covers a list of operator obligations: those of providers, authorised representatives, importers and distributors, of deployers of high-risk systems under Article 26, of notified bodies, and the transparency duties of providers and deployers under Article 50. The ceiling is €15 million or 3% of worldwide annual turnover, whichever is higher.

Tier 3 targets supplying incorrect, incomplete, or misleading information to notified bodies or national authorities in reply to a request. The ceiling is €7.5 million or 1% of worldwide annual turnover, whichever is higher.

Obligations outside these lists, such as AI literacy under Article 4, have no EU fine ceiling. Article 99(1) leaves the penalties for those to national law, which may include fines, warnings and non-monetary measures.

The SME Rule: the Lower Figure Applies

For large companies, "whichever is higher" means the percentage bites once turnover is high enough. For SMEs and start-ups it works the other way round. Article 99(6) says that for SMEs, including start-ups, each fine is capped at the percentage or the amount, whichever is lower. This is not a discretionary reduction you have to argue for. It is the ceiling.

So a company with €10 million in annual revenue that commits a Tier 2 violation faces a maximum of 3% of turnover, or €300,000, not €15 million. For a Tier 1 violation the maximum is 7%, or €700,000.

The Digital Omnibus on AI (Regulation (EU) 2026/1744) added a similar rule for small mid-cap enterprises (SMCs): for them the lower figure applies to Tier 2 and Tier 3 fines, but not to Tier 1 prohibited practices (Article 99(6a)). Article 99(1) also requires Member States to take the interests and economic viability of SMEs and SMCs into account when imposing penalties.

How the Amount Is Set Within the Ceiling

The ceilings are maximums. Within them, Article 99(7) lists what must be taken into account, as appropriate, when deciding whether to fine and how much. The factors include:

  1. The nature, gravity and duration of the infringement and its consequences, including the purpose of the AI system and, where appropriate, the number of people affected and the harm they suffered.

  2. The size, annual turnover and market share of the operator.

  3. The degree of cooperation with the national authorities to remedy the infringement and mitigate its effects.

  4. The degree of responsibility, taking into account the technical and organisational measures the operator had implemented.

  5. The manner in which the infringement became known, in particular whether the operator notified it.

  6. The intentional or negligent character of the infringement.

  7. Any action taken to mitigate the harm suffered by affected persons.

  8. Any other aggravating or mitigating factor, such as financial benefits gained or losses avoided.

Duration matters more than most deployers expect. Running a non-compliant system for 14 months weighs differently from running it for 6 weeks before correcting course.

Factors That Push Fines Up

Read against Article 99(7), these circumstances point toward the higher end of the range:

  • Intentional conduct. Deliberately disabling a human oversight mechanism weighs far heavier than a configuration error.

  • Poor cooperation. Stonewalling investigators, providing incomplete documentation, or delaying responses works against you, and misleading information is a separate Tier 3 infringement.

  • Serious harm or many people affected. A recruitment AI that screened 50,000 candidates unlawfully is a different case from one that screened 500.

  • Financial benefit. Gains made through the infringement are an explicit factor.

Factors That Push Fines Down

The same list cuts the other way. The most useful mitigating factors are:

  • Proactive disclosure. Reporting a problem before the authority discovers it is an explicit factor.

  • Immediate corrective action. Disabling the non-compliant system the moment a problem is identified, with documented timestamps, shows you acted to mitigate harm.

  • Existing compliance infrastructure. A written AI governance policy, staff training records, and a system inventory are the technical and organisational measures Article 99(7) refers to.

  • Full cooperation during investigation. Providing complete documentation promptly and making staff available each contribute.

  • Limited actual harm. If no individual suffered a concrete negative outcome, that counts in your favour.

This is where the preparation you do now has a direct monetary value. A compliance officer who can produce a dated training log, a risk classification decision, and a supplier contract with responsibilities assigned is in a structurally different position than one who cannot.

What Enforcement Looks Like Today

National market surveillance authorities supervise the AI Act from 2 August 2026. The high-risk obligations for Annex III systems apply from 2 December 2027, and for AI in products under Annex I from 2 August 2028, so Tier 2 fines for high-risk deployers cannot arise before then. Prohibited practices (two new ones from 2 December 2026), Article 50 transparency and AI literacy already apply.

In the Netherlands, the implementing law is still a draft and the Autoriteit Persoonsgegevens has no AI Act powers yet; it is the intended coordinating supervisor. The draft provides fines for prohibited practices, high-risk and transparency duties and misleading information, but not for Article 4: a shortfall in AI literacy would lead to an order to fix it, backed by a penalty payment (last onder dwangsom). We are not aware of any AI Act fine on a company in the Netherlands so far.

The One Number You Can Control

Every factor that drives fines downward (proactive disclosure, cooperation, documented governance, limited harm) is something you build before an investigation starts, not during it. The calculation rewards organisations that treat compliance as an ongoing operational function rather than a crisis response.

Start with knowing what AI systems you deploy and how they are classified. A system inventory, a risk tier for each tool, and a clear record of who is responsible for oversight cost almost nothing to build. They are the foundation every other mitigating factor rests on.

If you are not certain how your current AI deployments would be classified under Annex III or what obligations attach to them, the two-minute compliance check at comply.khairos.ai gives you a starting point, and the beginning of the documentation trail an authority looks for.

References

  1. Article 99

  2. Regulation (EU) 2026/1744 (Digital Omnibus on AI)

  3. Annex III

  4. comply.khairos.ai

Not sure where your company stands?

Answer 15 questions about how your team uses AI. You get a compliance score and the gaps to close first, as a PDF.

Am I compliant? Check in 2 minHow Khairos Comply works