Methodology | ModelCompliance.ai

Methodology

ModelCompliance.ai reviews public documentation from B2B AI SaaS companies to understand how clearly they disclose AI-related governance, model use, customer-data handling, transparency, and buyer-facing evidence.

Important context

This is a public documentation/transparency review, not a legal compliance determination, legal opinion, certification, audit, or assurance engagement.

What we review

Public documentation, policies, trust centres, model/provider references, and related public materials.

What we do not claim

We do not assess private controls, legal status, or buyer-specific product deployments.

Evidence statuses

Found

Clear public evidence was found.

Partial

Related evidence was found, but it does not fully answer the check.

Not found

No public evidence was found during review.

Unclear

Evidence exists but is ambiguous or difficult to interpret.

Not applicable

The check appears not applicable based on the public product profile.

Initial evidence checks

AI usage disclosure

Model/provider disclosure

Customer data training statement

AI-specific privacy/data handling statement

Subprocessor or AI vendor disclosure

Human oversight statement

Model card/system card or equivalent

AI safety/testing/evals/red-teaming statement

AI incident response statement

Responsible AI/governance policy

Security/trust centre availability

Framework references or governance alignment

Framework-informed themes

Checks may be tagged to themes inspired by EU AI Act, NIST AI RMF / GenAI Profile, and ISO/IEC 42001. The mapping is thematic only and does not claim formal conformity, certification, legal status, or official endorsement.

Corrections and disputes

Submit evidence

Companies may submit corrections or additional public evidence for free.

Editorial review

Profiles are updated only when supporting public evidence justifies the change.

Under review

Disputed profiles can be reviewed internally and marked under review when appropriate.

Payment does not directly alter public profile scores or evidence statuses.

Evidence-pack boundary

Evidence-pack requests are documentation/readiness support. They do not purchase a score change, legal opinion, audit conclusion, certification, or publication outcome.