SWT3 Witness Protocol

Cryptographic accountability for AI systems. Open specification. Independent verification.

$pip install swt3-ai $npm install @tenova/swt3-ai $cargo add swt3-ai $dotnet add package swt3-ai $gem install swt3-ai
$python -m swt3_ai.demo $npx swt3-demo

Run a local demo in 10 seconds. No API keys. No signup.

280
Procedures
77
Frameworks
46,000+
Downloads
69
Countries
10
Languages

SWT3 (Sovereign Witness Traceability) is an open protocol for cryptographically anchoring evidence to immutable witness records while preserving data sovereignty. Any party can independently verify that evidence has not been altered since the moment it was witnessed, even after the raw source data has been cleared.

Integrations

OpenAI Anthropic AWS Bedrock LiteLLM Ollama vLLM Cohere LangChain LangGraph Google ADK CrewAI Vercel AI SDK MCP Qdrant Core ML Secure Enclave

How It Works

Four phases. No agents installed. No data retained. No central authority.

01

Witness

The SDK observes AI inference at the point of execution. Evidence factors are captured and SHA-256 hashed locally. Raw prompts and responses never leave your infrastructure.

02

Anchor

Factors are sealed into an SWT3 Witness Anchor, a tamper-evident receipt binding evidence to a verdict at a specific moment in time. The fingerprint formula is locked and identical across 10 languages.

03

Verify

Any party can independently verify the anchor using the open-source library or a browser. No API keys. No vendor access. SHA-256 runs locally. Just math.

04

Clear

The clearing engine strips sensitive content before it leaves your jurisdiction. The cryptographic proof survives. The underlying data does not. 100% integrity and 100% privacy simultaneously.

Anchor Format

SWT3-E-VULTR-NET-SC76-PASS-1773316622-96b7d56c0245
SegmentDescriptionExample
SWT3Protocol identifierSWT3
TIERDeployment tierE (Enclave), S (SaaS), H (Hybrid)
PROVIDERInfrastructure providerVULTR, AWS, AZURE, GCP
UCTControl categoryACC, AUD, CFG, IDN, NET, INT, AI, POL
PROCEDUREProcedure identifierSC76, AI-INF.1
VERDICTDetermination resultPASS or FAIL
EPOCHUnix timestamp (seconds)1773316622
FINGERPRINTTruncated SHA-256 (12 hex)96b7d56c0245

Fingerprint Formula

Canonical Fingerprint (locked) SHA256('WITNESS:{tenant}:{proc}:{fa}:{fb}:{fc}:{ts_ms}').hex()[:12]

The fingerprint formula is locked. Cross-language parity is verified across Python, TypeScript, Rust, C#, Ruby, Swift, and Go using 13 conformance test vectors.

Verification in Practice

1. Given anchor
SWT3-E-VULTR-AI-INF1-PASS-1773316622000-96b7d56c0245
2. Re-derive fingerprint from original factors
SHA256("WITNESS:VULTR_ENCLAVE:AI-INF.1:gpt-4o:1247:0.92:1773316622000")
= "96b7d56c0245..."
3. Compare
Anchor: 96b7d56c0245 Derived: 96b7d56c0245 MATCH

Any party with the original factors can independently verify the anchor. No API call required. No vendor contact. The math is the proof. Try it live.

Clearing Engine

Cross-border trust requires cross-border privacy. The clearing engine strips sensitive content before it leaves your jurisdiction. The proof survives. The data does not. Anchors are tenant-scoped and cryptographically isolated -- no tenant can read, query, or infer another tenant's evidence.

L0

Analytics

Full context: hashes, factors, model ID, provider, guardrails. For internal R&D and pre-deployment testing.

L1

Standard

Default. Hashes and factors only. No raw prompts or responses cross the wire. Production-grade privacy.

L2

Sensitive

Hashes, factors, model ID only. No provider metadata. For healthcare, legal, and PII workloads.

L3

Classified

Numeric factors only. Model ID hashed. Zero metadata. For defense, sovereign cloud, and air-gapped environments.

Witness Middleware

Transport-level witnessing with zero application code changes. Drop it into any HTTP framework or MCP server.

Zero-Code Integration

The Witness Middleware wraps AI tool-call transports at the HTTP layer. It intercepts requests and responses, computes witness anchors, and forwards traffic without modifying payloads. Works with Express, Fastify, Rails, Django, or any HTTP framework, including MCP (Model Context Protocol) servers. Every tool call is attested before it reaches your application code.

// TypeScript -- Express middleware (one line)
app.use(witnessMiddleware({ endpoint: 'https://api.example.com/witness' }));
# Python -- ASGI/WSGI middleware (one line)
app = WitnessMiddleware(app, endpoint="https://api.example.com/witness")

The middleware observes. It does not modify, block, or filter traffic. Witnessing is passive by default. Enforcement is opt-in via policy-as-code declarations in .swt3.yaml.

Regulatory Coverage

Native mapping to 77 frameworks. Not a crosswalk bolted on after the fact.

EU AI Act NIST AI RMF NIST 800-53 CMMC v2.0 FedRAMP DoD RMF NIST 800-171 SR 11-7 HIPAA / 21 CFR GPAI Code of Practice IMDA Agentic AI Japan AI Promotion Korea AI Basic Act ISO 42001 EU CRA NIS-2 EU Battery Regulation ASL-4 (Anthropic RSP) OWASP MCP Top 10

Browse all 280 procedures across 77 frameworks

Evidence exports in NIST OSCAL (SAR/SAP), W3C Verifiable Credentials, and machine-readable JSON.

For Auditors and Assessors: Map SWT3 procedures to your framework's assessment objectives. Browse the registry by framework, export evidence chains, and verify anchors independently.

Assessment Mapping Live Audit Portal

SDKs

Read the Spec Browse the Registry Test Vectors Verify an Anchor