Standards · alignment in progress
How Hlido aligns with emerging AI-assurance standards.
Hlido is an independent rating brand — not a notified body or certifier. As public AI-assurance standards mature, we align our taxonomy and process vocabulary to map to them. This is alignment in principle: the standard texts are still being finalized, so the crosswalk below maps to well-established AI-QMS and continuous-auditing themes and will be tightened to exact clauses as they land.
The two standards
- prEN 18286 — the European standard for an AI Quality Management System (process discipline: documentation, traceability, impartiality, evidence, corrective action). Maps to how Hlido runs the review process. (At enquiry stage; expected to finalize around Q4 2026.)
- ETSI TS 104 008 — continuous AI auditing. Maps to Hlido's cadence: we re-test on release, spec change, and freshness expiry rather than issuing a one-shot verdict. Continuous auditing is a genuine differentiator versus one-time audits.
Crosswalk A — scoring dimensions ↔ QMS / audit themes
| Hlido dimension | AI-QMS theme (prEN 18286-class) | ETSI continuous-audit principle |
|---|---|---|
| Product clarity | Documentation & transparency of the system assessed | Auditable description, kept current |
| Feature depth | Scope characterization / capability evidence | Re-characterize on version change |
| Trust signals | Provenance, security, integrity controls | Continuous integrity monitoring |
| Claim verification | Evidence collection & verification; traceability | Continuous claim re-validation (drift) |
| User experience | Performance evaluation under realistic conditions | Periodic performance re-test |
| Transparency & access | Openness / disclosure obligations | Ongoing disclosure currency |
| Editorial layer (the verdict) | Assessor competence + impartiality | Findings refreshed on cadence |
Crosswalk B — process steps ↔ standard vocabulary
| Hlido step | QMS / audit term |
|---|---|
| Intake (claims, pricing, docs, MCP card) | System characterization / assessment scoping |
| Tier classification (+ public exclusions) | Assessability determination + documented exclusions |
| Claim audit (PASS / FAIL / UNVERIFIED + evidence) | Evidence collection + verification + traceability |
| Behavioral test (fixed suite, isolated, graded) | Performance evaluation under controlled conditions |
| Evidence capture (C2PA-signed / terminal cast) | Audit record / tamper-evident evidence |
| Signed verdict + claim-vs-evidence table | Assessment report / audit finding |
| Re-test on release / spec / freshness | Continuous auditing (ETSI TS 104 008 core) |
| Right of reply | Impartiality / appeals mechanism |
| Zero per-agent customization | Consistency / non-discrimination of method |
Scope — what this is and isn't
- Is: a transparency statement of how our independent method maps to recognized assurance themes, refined as standards finalize.
- Is not: a certification, a notified-body assessment, or a compliance claim. We do not sell a license or a government stamp, and we do not pursue accreditation.
- Unchanged: the exact scoring-rubric weights remain private (on a timer). Alignment is about taxonomy and process, not a rubric change.