For Regulators
Auditable to the
last assertion.
Aviation certifies against DO-178C. Nuclear certifies against IEC 61508. Clinical AI has had no quantitative equivalent — until now. The methodology is published, the audit trail is complete, and the standard is free for any authority to reference. The same de-identified infrastructure additionally supports privacy-preserving syndromic early-warning for public-health authorities — an outbreak signal produced without ever exporting patient identity, aligned to the IHR (2005) and specified in WP-007.
Quantitative bound, IEC-grade methodology
Fault Tree Analysis per IEC 61025 over 23 independent barriers; worst-case 3.23 × 10⁻⁶ per assertion,¹ with stated assumptions and degraded-mode analysis.
Maps to EU AI Act Articles 9–17
Risk management, data governance, human oversight, accuracy, and quality management — each requirement answered by a named, auditable mechanism.
100% pipeline audit coverage
Every action, decision, and data transformation cryptographically sealed and independently verifiable by third parties — including against the operator.
Uncertainty declared, never hidden
A formalized non-answer protocol outputs the disagreement, a hypothesis, and resolving diagnostics — the failure mode your guidance documents ask for.
An open, royalty-free standard — DSS
Four certification levels with explicit probability thresholds, adoptable by any authority, any vendor, any procurement framework. No licensing. Attribution only.
FDA Q-Submission engaged
Pre-submission dialogue on quantitative certification of verification architectures is underway. EU and national authorities: the same briefing is open to you.
¹ Fault Tree Analysis per IEC 61025, common-cause adjusted, worst-case operational degradation — WP-001. SIL equivalence refers to the demand-mode numerical target only, as stated in the paper.