How We Verify Our Claims

Every performance claim is backed by cryptographic proof.
Don't trust us. Verify independently.

≤ 20ms
Policy p99
99th percentile latency
≤ 800ms
E2E p95
95th percentile end-to-end
≥ 98%
Evidence Health
Attestation success rate

Independent Verification

Download our verification scripts to independently test our claims. All scripts are open source and cryptographically signed.

Verify Trust Receipts (ATR)

Validate Ed25519 signatures on AI Trust Receipts

Download
curl -fsS https://start.railguard.ai/trust/verify-atr.sh | bash
Verifies: Signature authenticity, ATR schema compliance, timestamp freshness

Verify Latency Claims

Run 100 requests and measure p99 latency

Download
curl -fsS https://start.railguard.ai/trust/verify-latency.sh | THRESHOLD_MS=3.0 bash
Verifies: Policy evaluation latency, E2E request latency, SLO compliance

Verify All Claims

Complete verification suite (ATR + latency + health)

Download
curl -fsS https://start.railguard.ai/trust/verify.sh | bash
Verifies: All SLO metrics, signature validity, evidence completeness

How It Works

1

Daily Snapshots

Every day at 2:15 AM UTC, GitHub Actions fetches live metrics from our production API endpoint (/api/status/dials).

2

Cryptographic Signing

Metrics are signed with our Ed25519 private key (256-bit elliptic curve). The public key is published at /trust/public-key.pem.

3

Public Distribution

Signed receipts are uploaded to S3/CloudFront and served at /trust/YYYY/MM/DD/performance.json. You can verify any past snapshot.

4

Independent Verification

Download our verification scripts (above) to check signatures, replay latency tests, and confirm SLO compliance from your own machine.

AI Trust Receipts (ATR) v0.1

We're pioneering the AI Trust Receipt (ATR) specification — a standardized format for cryptographically verifiable evidence of AI governance.

ATR Schema Example

{
  "version": "0.1.0",
  "timestamp": "2025-01-30T14:23:17Z",
  "metrics": {
    "policy_p99_ms": 18,
    "e2e_p95_ms": 742,
    "evidence_health_pct": 99.1
  },
  "signature": {
    "algorithm": "ed25519",
    "value": "abc123...",
    "public_key_url": "https://start.railguard.ai/trust/public-key.pem"
  },
  "issuer": "Railguard AI Gateway",
  "receipt_id": "atr-2025-01-30-142317"
}

Why this matters: Anyone can claim their AI is safe and fast. Only cryptographic receipts can prove it.

Challenge to Competitors

We challenge Guardrails AI, Lakera, Arthur AI, and Aporia to publish:

We'll wait. Until then, we're the only AI governance platform where "trust" is more than marketing — it's cryptographically provable.

Explore the Full Trust Center

Browse all historical snapshots, download raw data, and see our live Grafana dashboard.

Trust CenterCompare vs. Competitors

Frequently Asked Questions

Why should I verify your claims?

Because anyone can say they're fast and secure. Independent verification removes the "trust us" requirement. Run the scripts yourself and see the evidence.

How do I know the scripts aren't rigged?

All scripts are open source and MIT-licensed. You can read the code, modify it, or write your own. The public Ed25519 key is published at /trust/public-key.pem and has been the same since launch.

Can I verify past snapshots?

Yes. All snapshots are archived at /trust/YYYY/MM/DD/performance.json. You can verify any historical date.

What if my verification fails?

If signature verification fails, it means the receipt has been tampered with (extremely unlikely). If latency tests fail, it means we're not meeting our SLOs (we'll know because our CI gates will have failed). Either way, contact us immediately.

Do competitors offer verification scripts?

Not yet. See our competitive comparison for details.

Ready to See It in Action?

Start with a free trial. Provision your gateway in 15 seconds with contactless self-service.

Try Railguard Free
How We Verify | Railguard | Railguard AI