The Regulatory Sandbox Coordinator manages the organisation's participation in AI regulatory sandboxes (Arts. 57–59) and testing of high-risk AI systems in real-world conditions (Arts. 60–61, or within a sandbox under Art. 58(4)). Role 1.1 Directs which innovations proceed. Role 3.2 owns isolated security-testing environments. Role 1.4 files risk. This role prepares sandbox applications and plans, makes sure safeguards, oversight and data protection agreed with the competent authority are in place, runs the reporting, and returns results to Implement without letting sandbox status stand in for conformity.

Standing duties

• Prepare the sandbox plan agreed with the competent authority (Art. 57(5)) and real-world testing plans submitted to, and approved by, the market surveillance authority (Art. 60(4)(a)–(b)): scope, safeguards, oversight, data protection, exit criteria.
• Recommend at testing start, to Role 1.2 (sole authority contact; no veto, ROLES13 t269): no sandbox or real-world test of a high-risk or agentic system without an agreed plan, named overseer (Art. 60(4)(j)), registration where required (Art. 60(4)(c)), informed consent (Art. 61), reversibility of outputs (Art. 60(4)(k)) and isolation arrangements confirmed by Role 3.2.
• Run reporting to the authority — serious incidents in real-world testing under Art. 73 (Art. 60(7)); suspension, termination and final outcomes (Art. 60(8)) — and return results, incidents and lessons to Roles 1.4, 16.1 and 22.1.
• Raise ACT02 when a test would exceed agreed scope or purpose; raise ACT03 on incidents during testing.
• Do not treat sandbox participation, written proof or an exit report (Art. 57(7)) as conformity, residual-risk acceptance or release approval; they are evidence that conformity assessment and market surveillance may take into account.

Accountable to: AI Governance Lead / Head of AI Innovation (or GM: Regulatory Affairs)

Primary domain: Establish (EST04 innovation) and Implement (sandbox and real-world testing plans). Does not Direct or accept residual risk.

Linked EN 18286 clauses: 4.3, 8.1.2, 8.5

Format: 3 hours of training, EUR 200.