aiegis in the Irish AI Regulatory Sandbox Framework
15-layer security platform built, tested (200+ pytest files across api_v2 + registry + grid), and deployed. Live at aiegis.ie. Open-source checker published. SDK available. No R&D needed — deployment-ready today.
Submit to National AI Office for sandbox participation. Demonstrate live platform capabilities. Propose evaluation criteria with 2-3 NCAs (Central Bank, DPC, Coimisiun na Mean).
Deploy aiegis in sandbox environment under NCA supervision. Monitor real AI agents in controlled setting. Validate compliance classification accuracy. Collect performance data: latency, accuracy, false positive rates. Generate evaluation report for each participating NCA.
Full rollout to participating NCAs based on sandbox results. Formal procurement via eTenders. Expand to additional NCAs. Ongoing monitoring and updates aligned with evolving EU AI Act requirements.
Proposed metrics for sandbox evaluation, aligned with Head 35 requirements:
| Metric | Target | Current |
| Risk classification accuracy | >95% | 100% (32/32 test cases) |
| Adversarial resistance | >90% blocked | 100% (28/28 blocked) |
| EU AI Act article coverage | All high-risk | 9/9 (Art. 9-15, 72) |
| OWASP AI coverage | >80% | 10/10 (100%) |
| Uptime | >99.5% | 99.9% (since deployment) |