How to See What Your Auditor Receives from SWT3 Witness Anchors

You instrument AI systems with the SWT3 SDK. Your auditor verifies the evidence those instruments produce. This guide shows both sides of the same cryptographic artifact, so you know exactly what your auditor receives, how they verify it, and which regulatory requirements your instrumentation satisfies.
Universal protocol. SWT3 is an industry-agnostic cryptographic witness protocol. The evidence described in this guide is identical regardless of your AI system's domain, your auditor's framework, or your deployment infrastructure.
Contents
  1. Two Sides of One Anchor
  2. The Auditor's Workflow
  3. What Maps Where
  4. The Anchor Anatomy
  5. Your Auditor's Guides
  6. Links

1. Two Sides of One Anchor

When you instrument your AI client with the SWT3 SDK, every inference generates a witness anchor. This anchor is a single cryptographic artifact with two audiences.

You (the developer)

"My system is instrumented. Evidence is being generated automatically. Every inference, every tool call, every access decision is recorded with a tamper-evident fingerprint."

Your auditor

"I have a tamper-evident record I can verify independently without trusting the developer. The SHA-256 fingerprint either matches the evidence factors or it does not."

The anchor does not care who reads it. The SHA-256 fingerprint either matches the evidence factors or it does not. This is what makes SWT3 a protocol, not a platform. Verification requires nothing but the hash function.

2. The Auditor's Workflow

Here is what happens, step by step, when an auditor evaluates your SWT3 evidence.

1

Receives a read-only portal link

Your organization generates an audit share link. The auditor opens it in any browser. No account, no installation, no vendor relationship required.

2

Sees every procedure with a verdict

The portal displays a table of all instrumented procedures. Each row shows the procedure ID, the title, the verdict (PASS, FAIL, or INHERITED), the last witnessed timestamp, and a verify link.

3

Examines evidence factors on hover

Hovering over any procedure reveals the three evidence factors: the standard being evaluated, the evidence collected, and the context of the evaluation. Framework badges show which regulations this procedure satisfies.

4

Verifies any anchor independently

Clicking "Verify" opens the public verifier at /verify. The auditor enters the anchor's fingerprint. The verifier re-derives the SHA-256 hash from the factors and confirms authenticity. No server, no API key, no trust assumption.

5

Prints the assessment checklist

The auditor navigates to the assessment checklist page, selects their framework, and generates a printable checklist. Every procedure you instrumented appears as a row they can check off.

6

Files the evidence

The auditor now has verified anchors, a signed checklist, and a framework-specific walkthrough guide. This is stronger evidence than any attestation letter because it is independently verifiable and tamper-evident.

3. What Maps Where

The following table shows how your SDK instrumentation maps to the evidence your auditor verifies and the regulatory requirements it satisfies.

What You Instrument What Your Auditor Verifies EU AI Act NIST AI RMF
Client wrapping AI-INF.1 Inference witnessed with timestamp and model ID Art. 12(1) GOVERN 2.1
RAG provenance AI-RAG.1 Document hashes and retrieval metadata Art. 13 MAP 2.3
Tool witnessing AI-TOOL.1 Tool call chain with execution timestamps Art. 12(1) GOVERN 1.7
Agent identity AI-ID.1 Agent fingerprint across all anchors Art. 12(1) MAP 1.1
Access control gate AI-ACC.1 Authorization decision per inference Art. 9(4)(c) MANAGE 2.4
Anchor revocation AI-REV.1 Revocation with reason code linked to original Art. 14(4)(d) MANAGE 2.4
Model metadata AI-MDL.1/5/6 Version, weights hash, adapter stack Art. 11 GOVERN 1.2
Drift detection AI-DRIFT.1 Model output deviation from baseline Art. 9(2)(b) MEASURE 2.6

View the complete mapping for any framework at the Assessment Mapping page.

4. The Anchor Anatomy

Every SWT3 Witness Anchor is a single string with seven segments. Each segment carries specific meaning for both developer and auditor.

SWT3-E-AWS-AI-INF1-PASS-1780752000-a3f7c21d9b04
Protocol + Tier
Provider
UCT Domain
Procedure
Verdict
Epoch
Fingerprint
You generated this

Your SDK call produced this anchor. The fingerprint (last 12 hex characters) is derived from SHA-256 of the evidence factors you provided: the standard evaluated, the evidence collected, and the context of the evaluation.

Your auditor verifies this

The auditor enters this anchor at /verify. The verifier re-derives the fingerprint from the same factors. If the derived fingerprint matches the anchor, the evidence is authentic and has not been modified since it was witnessed.

5. Your Auditor's Guides

Your auditor follows a framework-specific walkthrough guide that maps these procedures to their regulatory requirements. Each guide shows the auditor exactly what to check, what a PASS verdict means, and what common findings look like.

EU AI Act Walkthrough
16 article groups, 80 procedures. Covers Articles 9 through 72 with SWT3 evidence mapping for each obligation.
NIST AI RMF Walkthrough
4 functions (Govern, Map, Measure, Manage), 80 procedures. Maps every subcategory to SWT3 evidence.
CMMC Walkthrough
12 practice families, 30 procedures. Aligns SWT3 evidence with CMMC Level 2 practices for AI in defense contracting.
SR 11-7 Walkthrough
6 sections, 19 procedures. Maps SWT3 evidence to Federal Reserve model risk management expectations.

Explore the evidence chain