SANCTIAGENT · RUBRIC REGISTER

Claims & methodology

Rubrics are versioned self-assessments. Scores indicate documentary control readiness; they are not independent assurance, legal opinions, insurance advice, or compliance conclusions.

Regulatory and insurance claim rows are awaiting qualified human review. Do not rely on this register as legal advice.

Instrument status

Attribution mapping

Article 12 concerns logs, Article 14 human-oversight design, Article 26 deployer duties, Article 50 transparency, and Articles 72–73 post-market monitoring and serious incidents. These provisions contribute evidence; they do not alone decide civil liability.

The source register, retrieval dates, review dates, and outstanding sign-off status are recorded in tasks/claims-register.csv. Legal and insurance review remains required before commercial launch.

Optional BYOK assistant (capability disclosure)

The optional draft assistant runs entirely in the user's browser: prompts and self-assessment context are sent only to the inference provider the user selects, using an API key stored in that browser; SanctiAgent servers never receive the key, the prompt or the draft output (CL-027).

Model output is an unverified working draft for human review — not legal, insurance or compliance advice — and the deterministic scorecard remains the source of truth (CL-028). Free-key signup pages for the six listed providers appear in the assistant panel on every tool page; free-tier quotas and terms belong to those providers and can change without notice (CL-029).