QgenticQgentic
Qgentic / platforms / ai act
Qgentic AI Act · European Union

The EU AI Act register, classified by the Act's own tests.

Qgentic AI Act turns the records you already hold into a structured register, with each risk category computed by code from the system's declared capabilities and the Act itself. The UK's own instruments run on the same platform.

Who it is for

EU providers and deployers

Organisations placing or deploying AI systems in the EU: transparency duties, the high-risk regime, Annex IV documentation and incident reporting timelines.

Non-EU providers selling into the EU

Organisations outside the EU whose AI systems are placed on the EU market or whose output is used within it. The Act reaches them, and the register does too.

UK firms and public bodies

The UK has no AI Act. PRA SS1/23 model inventories and ATRS transparency records are covered by a separate module on the same platform: Qgentic AI Governance (UK).

The rulebooks it covers

EU · Regulation (EU) 2024/1689

✓ EU AI Act register

Each AI system is a structured record. Code computes its risk class, checks its Annex IV documentation is complete and runs its incident reporting clocks, and the validation engine blocks a configuration that breaks the rules.

UK · Regulatory principles

UK instruments have their own page

The UK deliberately has no AI Act, so its instruments stand on their own. PRA SS1/23 model inventories and ATRS public records are covered by Qgentic AI Governance (UK), on the same engine, dashboard and licence.

How a system gets classified

  1. IntakeSystem metadata arrives by file upload, CSV or REST API.
  2. ClassificationThe risk category is calculated from the Act's own tests (Articles 5 and 6, Annex III). A manual classification that contradicts them is rejected.
  3. ValidationThe engine checks documentation completeness, governance checkpoints, incident reporting timers and ATRS field requirements.
  4. MonitoringAlerts go out for missing documentation and upcoming incident reporting deadlines.
  5. Approval and exportA named person reviews and approves the data, and the cryptographically hashed filing package is generated.
How an AI system is classified and its technical file assembled Your model inventory, technical files and deployment context feed the Qgentic engine. The model reads the model cards and drafts Annex IV sections; code decides the classification: Article 5 prohibitions, the Annex III high-risk tests, the GPAI compute threshold and the fundamental-rights impact assessment trigger. Classification is a test rather than a guess: a class you assert that the Act's own tests contradict is rejected, and every answer names the article it came from. The run then stops at a named person approver, whose name, note and timestamp are written to the audit chain, before the AI system register, Annex IV checklist, registration record and incident log are exported with a manifest and a SHA-256 hash for every file. § YOUR SYSTEMS § THE QGENTIC ENGINE § THE GATE § THE TECHNICAL FILE Model inventory owner · purpose · market 5 AI systems Technical files training + testing docs one per system Deployment context who it affects, where and on whose behalf MODEL READS CODE DECIDES reads the model cards extracts the purpose drafts Annex IV text summarises for review Article 5 prohibitions Annex III risk tests GPAI compute threshold FRIA trigger rules CLASSIFICATION IS A TEST A class the Act's own tests contradict is refused. AIACT-CA-008 · every answer names its article 5 systems · 2 high-risk · 0 prohibited AWAITING APPROVAL A named person signs the file off. marta.lindqvist approver · admin The name, the note and the timestamp go on the chain. no self-approval Annex IV, assembled ai-system-register.csv annex-iv-checklist.json registration-record.json incident-log.json manifest.json SHA-256 per file the Article 73 clock starts on the incident audit chains VALID Regulation (EU) 2024/1689: the same register carries PRA SS1/23 model inventories and ATRS records, because underneath them it is one inventory.
Risk class is where an AI register is usually softest, so code decides it: prohibited, high-risk, GPAI and the FRIA trigger are each settled by the Act's own tests, and a classification you assert that those tests contradict is refused. Every answer carries the article it came from, which is what an Annex IV file has to survive.

What code decides, and what you supply

Decided by codeSupplied by your team
Risk category computation and enforcementSystem capabilities and deployment context
Annex IV and governance completeness checksContent of technical documentation
Incident reporting deadline calculationsIncident severity assessment
UK risk tier computation and validation trackingMateriality thresholds and validation schedules
Cryptographic audit logging and export generationFinal review and package approval

Every regulatory decision is made by deterministic code; no AI model takes part in governance or classification.

What you get

The export

For the EU, the AI-system register, Annex IV checklists, Annex VIII records and incident logs. For the UK, the model inventory, attestation summaries and ATRS records. Every file is hashed and exported after approval.

ai-system-register.csv, annex-iv-checklist.json, model-inventory.csv, manifest.json

Scope and limits

It validates general-purpose AI models and fundamental-rights impact assessments, and it prepares: submission to the EU database and conformity assessment go through your existing channels. Authorised representative requirements (Article 54) are out of scope for now.

What it costs

Pilot

£3,500 / €4,000 flat, 4-6 weeks

The first system inventory and a compliance gap analysis. The fee is credited to the annual licence.

SaaS

from £1,500 / €1,750 /month + usage

Hosted, and scaled to the number of AI systems you register.

ATRS Publisher

£190 /month per public body

For UK public bodies, with up to 10 transparency records.

On-premise

from £24K / €28K /year, Sealed Site

An annual licence to run air-gapped on your own infrastructure.

Every edition's price is on the pricing page. You can evaluate it offline on sample data first: ask for an evaluation build.

Frequently asked questions

What is the EU AI Act register?

A structured record of AI systems as mandated by Regulation (EU) 2024/1689. It includes system classifications, deployment contexts, and Annex IV technical documentation for high-risk systems.

How is an AI system's risk category decided?

Code computes it from Articles 5 and 6 and Annex III of the AI Act, and a manual override that breaks those rules is rejected.

Does the UK have an AI Act?

No. The UK relies on principles applied by its existing sector regulators, with no single statute. PRA SS1/23 model inventories and public-body ATRS records are covered by Qgentic AI Governance (UK).

What is SS1/23?

SS1/23 is a Bank of England supervisory statement requiring PRA-regulated entities to maintain a model inventory with risk tiering and validation schedules. It is served by Qgentic AI Governance (UK), not by this module.

Does this govern models or document them?

It documents and classifies. It does not inspect model weights or test model performance.

Are general-purpose AI models supported?

Yes. It computes the duties that apply under Articles 53 and 55, and handles Article 27 fundamental-rights impact assessments. Current status is in the product notes.

EU AI Act guides