The conformance report government, university, and EU buyers can trust.

SecondSigner is an evidence-first VPAT 2.5 / Section 508 ACR generator. Every WCAG 2.1 Level A/AA criterion carries Confirm / Correct / Mark-unknown evidence instead of a default “Supports”, and a DeafBlind evaluator personally co-signs the result. The no-fabrication doctrine, turned into a procurement-grade trust object.

See a sample report Open the authoring tool How it works

Built so a fabricated “Supports” cannot print

Pure function of evidence

Never fabricate

Conformance is derived from the evidence set. An unbacked claim is nulled back to the honest unknown default; a confident, fabricated “Supports” cannot reach the page.

A hard export gate

Honest unknown

“Not Evaluated” is AAA-only in the official template, so any Level A/AA row left unknown blocks finalization rather than printing a guess.

Identity, not assertion

Cryptographic co-signature

Every manual attestation and automated row carries an Ed25519 signature; finalization produces a real countersignature verified against the signer’s public key.

axe-core + AA contrast

Honest automation

Automated checks propose “Does Not Support” only from a concrete failure. A green run graduates nothing — absence of a detectable defect is not conformance.

Same gates, certified

They ship what they use

The authoring surface and the rendered report pass the same accessibility and brand gates this product certifies against.

ITI WCAG edition

Official VPAT 2.5 layout

Output is the ITI VPAT 2.5 (WCAG edition) structure — product information, applicable standards, the Terms, and the Level A / Level AA criteria tables.

Signing is a deliberate human gate

The export gate ships closed. SecondSigner refuses to produce a signed VPAT until the limitations language has had counsel review and the sole signer has applied a real countersignature against a specific evidence snapshot. The authoring tool here compiles the working draft; finalization is a separate, human-owned step.

Start a draft

Built with AI, reviewed by humans. Any AI-generated image carries a visible identifying mark on the image itself. No exceptions.