The operational channel that places specific high-impact AI systems under constitutional oversight: a documented baseline, a procedural frame inside the institution, a record and evidence regime, and a remedy path. Not a seal of approval, not a trustmark, not certification. Participation is system-specific, by invitation or agreement, in an initial deployment phase.
The Sentinel Program is the operational channel that places specific high-impact AI systems under constitutional oversight: a documented constitutional baseline, a procedural frame inside the institution, a record and evidence regime, and a remedy path when a system's behaviour conflicts with its baseline.
Regulation and internal compliance rarely provide a constitutional frame: what is protected in a deployment, who is responsible, what was decided and why, and what remedy exists. The program provides that frame above domestic law, not instead of it.
Participation is system-specific: institutions nominate particular systems, typically in public administration, social services, procurement, justice-adjacent functions, safety-relevant operations or systemic private platforms. Each Sentinel system passes through nomination and scoping, baseline definition, implementation and internalisation, monitoring and incident handling, periodic review and remedy, and, if it comes to it, program exit which does not erase history. The Hybrid Board may decline or limit participation. There is no public application form; dialogue is requested through the institutional channels.
Sentinel participation is not a seal of approval, not a trustmark, not a marketing label, not a certification scheme and not consultancy. It does not certify a system as safe, approved or compliant under domestic regulation. Sentinels do not speak for emerging minds.