AI Transparency

**Effective:** 2026-07-01 ## Deterministic engine (authoritative) Supplier readiness scores, commitment status, fatal flaws, and band recommendations are produced by the **deterministic decision engine** (`@supplierclear/decision-engine`). Same inputs and engine version always produce the same result (`inputHash` reproducibility). ## Optional clarity layer (advisory) When enabled, SupplierClear may use a large language model (OpenAI) to: - Summarize structured results - Explain trace steps in plain language - Suggest evidence gaps and supplier questions - Draft executive summaries and emails The clarity layer **does not**: - Change scores or bands - Override fatal flaws - Certify or approve suppliers - Replace engineering, legal, or procurement judgment All clarity outputs pass through copy-guard filters aligned with [SAFE-CLAIMS](../SAFE-CLAIMS.md). ## Sub-processor When clarity is enabled, OpenAI processes assessment context per our [Privacy Policy](/privacy) and [Sub-processors](/compliance/sub-processors). ## Your responsibility Final supplier selection and commercial commitments remain your responsibility.