Self-Promotion
Show off what you're working on
Start new thread

I rebuilt PolicyWatcher 4 around inspectable evidence, what would you audit first?

by•

Hi!

I’m Fabrizio, the maker of PolicyWatcher, an independent civic-tech project for monitoring changes in public company policies ->

For PolicyWatcher 4 Foundation Beta, I wanted to address a problem I kept seeing in policy-monitoring products: official text, detected changes, AI summaries, scores and final conclusions are often collapsed into a single output.

That makes the result difficult to inspect—and even harder to challenge or correct.

So we rebuilt the platform around a canonical evidence graph:

Entity → Document → Version → Change → Provision

The operating engine now follows six controlled stages:

Signal → Discover → Capture → Qualify → Publish → Correct

What changed in version 4:

• A Site Atlas mapping 23 product surfaces and 44 relationships.
• A Feature Atlas documenting 134 capability records across 15 domains.
• An adaptive public Evidence Console for citizens, GRC/legal teams, researchers and builders.
• Policy deep dives connecting official sources, captured versions, text differences and qualified AI-assisted summaries.
• Protected Admin operations across six responsibility areas and 19 consoles.
• Separate Admin and read-only Auditor presentations.
• A semantic evidence operation connecting an exact excerpt, evidence hash, canonical locator and review rationale.
• A 17-dimension Full‑V4 readiness evaluator with fail-closed publication controls.

The important design principle is:

A hash can prove integrity, but it cannot prove meaning. An AI summary can support review, but it is not clause evidence or legal advice. Publication eligibility does not certify a company or its compliance. The Full‑V4 taxonomy control is implemented in the current local v4 build, but it is not yet represented in the public Atlas registry or deployed to production.

The current audit reports 0/50 companies as Full‑V4 ready while accountable human review remains open.

I would genuinely value feedback from builders, researchers and GRC/legal professionals:

Which part would you audit first?

A. The canonical evidence chain
B. The taxonomy and readiness model
C. The public dashboard and policy deep dive
D. The Admin/Auditor control plane

And most importantly: are the boundaries between evidence, interpretation and publication clear enough?

A video of the platform:

Explore PolicyWatcher:

Thanks for taking a look and critical feedback is especially welcome!

4 views

Add a comment

Replies

Be the first to comment