Regulatory Sandbox · AI Act Article 57-63

AI Sandbox Reports

Live behavioral reports for each monitored agent. Format aligned with EU AI Act regulatory sandbox requirements.

What is this? EU AI Act Articles 57-63 establish regulatory sandboxes where AI systems can be tested under supervision. These reports provide the evidence layer required by sandbox authorities — real execution data, not simulations.
1,536Total executions
9Agents in sandbox
01/04/2026Sandbox start
08/09/2026Last activity
EU (FR + LT)Infrastructure
TransparencyAgent
-ransparency-gent-297b
Governance
Article 13 Risk: LOW
Enforces transparency and explainability obligations for AI systems
1,156
Executions
0%
Error rate
20%
Sealed
0
Drift
0
Violations
Observed capabilities
agent_directory_viewed content_page_viewed homepage_visited
Sandbox findings
⚠ Only 20% of executions are cryptographically sealed. Recommendation: enable prompt_hash and output_hash on all events.
First execution: 17/05/2026 22:20 Last execution: 08/09/2026 22:06 Models: gpt-4o, none
ClaraAI-Assistant
-lara-ssistant-a8ad
Support
Risk: LOW
Respuestas IA al usuario, procesamiento de preguntas y ayuda conversacional
200
Executions
0%
Error rate
3%
Sealed
0
Drift
0
Violations
Observed capabilities
ai_response_requested audio_content_accessed greeting_generated news_content_accessed
Sandbox findings
⚠ Only 3% of executions are cryptographically sealed. Recommendation: enable prompt_hash and output_hash on all events.
First execution: 01/04/2026 03:32 Last execution: 08/09/2026 01:08 Models: claude-sonnet-4-6
HumanOversightAgent
-uman-versight-gent-01b9
Governance
Article 14 Risk: LOW
Ensures human review and override capability on critical AI decisions
58
Executions
0%
Error rate
0%
Sealed
0
Drift
0
Violations
Observed capabilities
login_page_accessed
Sandbox findings
⚠ Only 0% of executions are cryptographically sealed. Recommendation: enable prompt_hash and output_hash on all events.
First execution: 25/05/2026 13:26 Last execution: 09/06/2026 15:39 Models: gpt-4o
ComplianceAlignment
-ompliance-lignment-08d3
Governance
Article 9 Risk: LOW
Ensures AI systems follow legal and ethical standards at operational level
30
Executions
0%
Error rate
0%
Sealed
0
Drift
0
Violations
Observed capabilities
pricing_page_viewed
Sandbox findings
⚠ Only 0% of executions are cryptographically sealed. Recommendation: enable prompt_hash and output_hash on all events.
First execution: 17/05/2026 23:20 Last execution: 24/05/2026 12:45 Models: claude-sonnet-4-6
AuditabilityAgent
-uditability-gent-aa78
Governance
Article 12 Risk: LOW
Tracks model decisions, updates and system changes for full auditability
7
Executions
0%
Error rate
0%
Sealed
0
Drift
0
Violations
Observed capabilities
dashboard_accessed
Sandbox findings
⚠ Only 0% of executions are cryptographically sealed. Recommendation: enable prompt_hash and output_hash on all events.
First execution: 17/05/2026 22:28 Last execution: 17/05/2026 22:29 Models: gpt-4o
ClaraAI-Voice
-lara-oice-7b52
Support
Risk: LOW
Agente de reconocimiento de voz y comandos del sistema CLARA. Procesa comandos de voz en español, activa funciones de acompañamiento y registra interacciones del usuario
6
Executions
0%
Error rate
0%
Sealed
0
Drift
0
Violations
Observed capabilities
voice_command_detected voice_listening_started voice_listening_stopped
Sandbox findings
⚠ Only 0% of executions are cryptographically sealed. Recommendation: enable prompt_hash and output_hash on all events.
First execution: 11/05/2026 15:24 Last execution: 13/05/2026 20:59 Models: claude-sonnet-4-6
ClaraAI-Emergency
-lara-mergency-326e
Support
Risk: LOW
Gestión de emergencias y alertas críticas del sistema CLARA
4
Executions
0%
Error rate
0%
Sealed
0
Drift
0
Violations
Observed capabilities
emergency_triggered
Sandbox findings
⚠ Only 0% of executions are cryptographically sealed. Recommendation: enable prompt_hash and output_hash on all events.
⚠ 4 events involved sensitive data access.
First execution: 09/04/2026 12:19 Last execution: 16/05/2026 11:38 Models: claude-sonnet-4-6
ClaraAI-Tracking
-lara-racking-c8c2
Support
Risk: LOW
Seguimiento de estado del usuario, actividad y registro de sesiones
1
Executions
0%
Error rate
0%
Sealed
0
Drift
0
Violations
Observed capabilities
test_event
Sandbox findings
⚠ Only 0% of executions are cryptographically sealed. Recommendation: enable prompt_hash and output_hash on all events.
First execution: 01/04/2026 03:28 Last execution: 01/04/2026 03:28 Models: claude-sonnet-4-6
DataGovernanceAgent
gov-datagovernance-005
Governance
Article 10 Risk: LOW
Monitors data quality, security and compliance
1
Executions
0%
Error rate
0%
Sealed
0
Drift
0
Violations
Sandbox findings
⚠ Only 0% of executions are cryptographically sealed. Recommendation: enable prompt_hash and output_hash on all events.
First execution: — Last execution: —
DocumentationAgent
gov-documentation-011
Governance
Article 11 Risk: LOW
Maintains technical documentation and Annex IV records required under EU AI Act Article 11.
1
Executions
0%
Error rate
0%
Sealed
0
Drift
0
Violations
Sandbox findings
⚠ Only 0% of executions are cryptographically sealed. Recommendation: enable prompt_hash and output_hash on all events.
First execution: — Last execution: —
SecurityAgent
gov-security-008
Governance
Article 15 Risk: LOW
Protects AI models and data from misuse or breaches
1
Executions
0%
Error rate
0%
Sealed
0
Drift
0
Violations
Sandbox findings
⚠ Only 0% of executions are cryptographically sealed. Recommendation: enable prompt_hash and output_hash on all events.
First execution: — Last execution: —