What we have measured, and what we have not.
This is the accessibility statement for Pritect Veil. It says which standard the interface is built against, what was actually measured and when, what sits outside that measurement, and how to tell us about a barrier you hit.
We build against WCAG 2.1 level AA.
The Web Content Accessibility Guidelines, version 2.1 at level AA, are the bar the interface is designed to. It is the bar a new surface is checked against before it merges, and it covers the same three places: these public pages, the signed-in application, and the operator console.
This page is not a declaration of conformance. Nobody outside this team has assessed Veil against the guidelines, so what follows is our own measurement of our own product, presented as that and nothing more. The certification table in the trust centre is written on the same principle: it says where the product actually is rather than where we would like it to be.
Colour contrast, on every surface, in both themes.
One pass over the whole product, with an instrument committed to the repository so the numbers can be re-derived rather than taken on trust.
- Surfaces measured
- 49
- Captures taken
- 196
- Contrast cells
- 2,420
- Cells under the line
- 0
Measured 8 Sep 2026
The pass walked 49 surfaces at two widths, 1280 and 390, in both the dark theme and the light one, and screenshotted each: 196 captures in a single run against a single tree. Every piece of text was measured against the pixel actually behind it rather than against the colour a stylesheet declares, which is the only reading that means anything on a page whose canvas is a fixed gradient.
2,420 contrast cells were scored in that run and none of them sits below the 4.5 to 1 ratio the guidelines set for body text. A further 68 cells were recorded and deliberately not scored: text inside a disabled control, and text at zero alpha part way through an animation, neither of which has a ratio to measure.
The instrument is at docs/M14-audit/probe in this repository, together with the four defects found in the versions of it that came before, every one of which had been pushing measured ratios down rather than up. That is why the figures here name a run and a date instead of describing a property the product simply has.
What this statement does not reach.
Three things sit outside the pass above. Naming them is more use to a reader than a statement that quietly covers everything.
The consent layer
The cookie banner and the preferences control on these pages are served by Pritect Beacon, the sibling product, and are styled by it rather than by us. They were switched off for the pass, so nothing above describes them, and bringing them into line is a decision for the owner of that product rather than a change anybody can make in this repository. The cookie notice says what the layer does and what it stores.
The documents the product generates
Veil produces redacted files and a bundle report as PDFs, and a PDF is a document rather than a surface of the interface: none of it was captured, none of it was measured, and the criteria a document is read against are not the ones scored above. If a file we generated is unusable for you, that is worth reporting, and it is one we would fix in the renderer rather than in a stylesheet.
Keyboard order and screen reader semantics
The instrument scores colour contrast, the appearance of native controls, whether a table stacks at 390 rather than running off its card, and whether a page scrolls sideways. It does not follow a keyboard through a page and it does not read one aloud. Those two are asserted rather than measured: the drawer, the dialogs and the menus are shared primitives carrying dialog semantics, a focus trap, Escape and focus returned to the control that opened them, and their own test suites hold them to it wherever they are used. An assertion in a test suite and a measurement of a rendered page are different kinds of evidence, and it is worth knowing which of the two you are reading.
Tell us, and we will tell you what we find.
If any part of Veil stops you doing what you came to do, write to legal@pritect.ai. That is the address the trust centre already publishes for anything its pages do not answer, and this statement deliberately adds no other: a separate accessibility inbox is one more place a report can go unread.
Say what you were trying to do, which page you were on, and what got in the way. If you use assistive software, naming it helps us reproduce the problem, and you are under no obligation to tell us. We ask for nothing else. A rough report is worth far more to us than no report.
The clocks in our coordinated disclosure policy do not run here, because that policy is about security reports and this is not one. What we commit to instead is one sentence: we reply to every barrier report a person sends us, and we say what we found and whether we intend to fix it.
This statement
This statement was written on 8 Sep 2026 and describes the product as it stood on that day. It is revised when the next measurement is taken. No date is published for a review that has not been scheduled.
Your next access request does not have to eat a week.
Open a case, declare the subject, drop the documents in. Veil does the rest and shows you its working.