Dashboard
Compliance
Document an AI system for EU AI Act style governance, including transparency, instructions for use, risk, data, evaluations, quality and incidents.
SDK / API and Hybrid systems have a compliance view next to the operations tabs. It collects the records an auditor or customer usually asks for.
| Tab | Purpose |
|---|---|
| AI Compliance Overview | Status of all compliance areas of the system |
| AI Transparency & Disclosure | Disclosure checklist and machine-readable transparency signals |
| AI Risk Assessment | Risk register of the system |
| Data Governance for AI | Datasets used by the system and their relevance |
| System Evaluation & Testing | Evaluations and their results |
| Quality & Governance System | Quality management records |
| Incident Management | Incidents and their handling |
| User Guidance & Documentation | Instructions for use for deployers |
| Technical Documentation & Compliance Artifacts | Technical documentation and exports |
AI Transparency & Disclosure
The AI Transparency & Disclosure Checklist records whether:
- users are informed they are interacting with AI,
- AI-generated content is labeled,
- deepfake or synthetic media is clearly disclosed.
Changes are saved automatically.
Inject machine-readable transparency metadata adds a transparency object to every Guard response of the system:
{
"transparency": {
"is_ai_generated": true,
"disclosure": "You are interacting with an AI.",
"article": "EU_AI_ACT_ARTICLE_50",
"injection_mode": "deterministic"
}
}Your application can use it to label AI output. The Node.js SDK also ships an AgentIDTransparencyBadge React component that logs proof that the disclosure was rendered; see Node.js SDK.
User Guidance & Documentation
The instructions for use are organized in eight sections:
- Identity & Contact of Provider
- Intended Purpose
- Accuracy, Robustness & Cybersecurity (and operating circumstances)
- Known/Foreseeable Risks and Misuse (health, safety, rights)
- Technical Capabilities Relevant for Explaining Outputs
- Target Groups / Persons
- Input Data Specifications (training/validation info)
- Deployer Guidelines for Interpreting Outputs
Fill them in manually or click Generate AI Draft, review the draft and click Save Instructions. You can Export Word (.doc) or create a Share Link: a secure read-only link that expires after 7 days. Sections that are still empty are marked Not provided yet in the export.
Human oversight
Assign who oversees the system in the Access tab. See Organization and access.
Conformity bundle
Export Conformity Bundle in the system header packages the system's evidence files for review. It is available to system admins and requires every evidence file to be locked first; otherwise the export stops and tells you how many files are still unlocked.