← Back to tools

Tool 09 · EU AI Act cockpit

Your AI system, under the European regime

A diagnosis of an AI system against Regulation (EU) 2024/1689: risk tier, obligations dated by your role, liability exposure and the insurance covers that respond. 100% in the browser, neither legal advice nor certification.

2024/1689
The EU AI Regulation, in force since 1 August 2024.
Official Journal of the EU
up to 7%
Of worldwide annual revenue: maximum fine for a prohibited practice.
AI Act, Art. 99
Aug 2026
Most of the regime applies, including Annex III high-risk systems.
AI Act, Art. 113
Annex III
Eight domains that class an AI system as high risk.
AI Act
Application timelineThe Regulation applies in phases. The deadlines that concern you are highlighted.
2 February 2025in force
2 August 2025in force
2 August 2026in force
2 August 2027in 319 d

AI system profile

Your role
Use domain
Model type
Interacts directly with people
Generates or manipulates content (text, image, voice)
Worldwide annual revenue€50M
High riskAnnex III · III.5
The most demanding regime

The use is listed in Annex III: the system is classed high risk, the fullest regime.

Verdict

A compliance program

High-risk system: the documentation base, human oversight and conformity assessment must be completed before the deadline. The liability exposure warrants a review of coverage.

Priority levers
  • Build the technical file and risk-management system without waiting for the deadline.
  • Formalize human oversight and traceability (logs, instructions, informing people).
  • Review E&O, product-liability and cyber policies to remove AI exclusions (silent AI).

Obligations & compliance0/6 met

Liability exposure82/100

LowHigh
Maximum fine at stake
€15M

For a breach of obligations: up to 3% of worldwide annual revenue or EUR 15M, whichever is higher.

The revised Product Liability Directive (2024) treats software and AI as a product and eases the claimant's burden of proof. The AI Liability Directive, however, was withdrawn from the Commission's work program in 2025: the fault regime remains national, a source of uncertainty to watch.

Insurability

Professional liability / E&OPartial
Product liabilityResponds
CyberPartial
Directors & officers (D&O)Responds
Regulatory finesGap

Regulatory fines (up to 7% of worldwide annual revenue) are in principle uninsurable; watch out too for silent AI, those traditional policies not designed for AI.

How it is calculated
Diagnosis parameters
RoleProvider
DomainCredit, insurance & essential services
ModelClassic AI system
Selected tierHigh risk
Sources & references
Based on Regulation (EU) 2024/1689 and Directive (EU) 2024/2853, plus:
EIOPA, supervision assurance UEsource →OCDE, politiques de l'IAsource →AI Incident Databasesource →AlgorithmWatchsource →
Assumptions & limits
  • Classification derived from use domain (Annex III), role and model type; it does not replace a case-by-case legal qualification.
  • A single system may carry several qualifications (e.g. GPAI embedded in a high-risk use).
  • Timeline per the Regulation: prohibitions Feb 2025, GPAI Aug 2025, high risk Aug 2026, embedded systems Aug 2027.
  • Liability: the revised Product Liability Directive is operative; the AI Liability Directive was withdrawn in 2025, leaving the fault regime to national law.
  • Indicative insurability map: the real response depends on exact policy wording and AI exclusions.
Go further
Price the exposure with the Agentic AI Pricer →

Educational and framing instrument. It is neither legal advice, nor a conformity assessment, nor a guarantee of insurability. The real classification of an AI system requires a case-by-case legal analysis and depends on evolving texts and guidance. Always check official sources and qualified counsel.