EU AI Act Control Mapping
EU
7 source references
4 mapped control areas
Regulated SaaS
About this template
A high risk deployment brings two jobs at once. Someone must show which obligations apply, and someone must show that the runtime actually behaves. This starter carries both halves. A reporting gap check names the articles in play, and a live chain governs each request. Neither half creates conformity, and the assessment work stays with your team.
The gap check for high risk systems is reporting only. It reviews the articles on risk management, data governance, documentation, record keeping, transparency, oversight and robustness. Findings come back through a separate report request. This part of the starter never blocks traffic and is not part of the enforcement chain.
What does run is a plain request and response chain. Injection screening rejects a prompt at 0.80, and identity must carry a user and an organization so a decision path has an actor. Every request must arrive with a value in the named consent header, or the gateway blocks it. Personal, sensitive, biometric and high risk decision data are the declared categories. The declaration is recorded with the outcome and never narrows the set of checked calls. Redaction masks national identifiers, bank account numbers, tax identifiers, health membership numbers and passports.
Language checks run in both directions and allow five deployment languages, so an answer outside that set is refused. A bias check also runs on output, but it fires only when a request mentions hiring, promotion or a performance review. High risk decision traffic outside employment therefore passes it untouched, so count it as monitoring and not as a fairness control. Citations must be present and must match the supplied source. A combined score of 0.80 and relevance of 0.85 are the passing marks, and anything under either one is blocked.
Chain enforcement is only one part of the obligation. Conformity assessment, technical documentation and post market monitoring are programs, not gateway settings. The audit marker only marks an evaluated request, and evidence storage and export stay with your team. Risk classification for your own system is the first decision, and this starter does not make it.
Source frameworks, standards and obligations
Control areas mapped by this template
Deployment options
Adapt the template in three steps
Review the scope
Confirm that the regions, source references, control areas and deployment options fit your use case.
Adapt the configuration
Open the template in the Verdictan console. Set its policies, thresholds, routing and review requirements for your environment.
Test before deployment
Test each policy outcome. Confirm that allowed, redacted, blocked and reviewed requests behave as expected.
Explore more Industry-Specific templates
Finance
Protect financial data, flag material nonpublic information, add supervisory notices and retain audit evidence with this starter.
Healthcare (U.S. — HIPAA)
Start with selected HIPAA controls for protected health information, minimum-necessary access, retention and clinical review.
Healthcare (EU — GDPR)
Map European health-data controls for GDPR Article 9 and EU AI Act risk management with this starter.