Public substrate sandbox

Try the TrustAI substrate live.
Sixty seconds, no procurement.

Pick one of three pre-canned scenarios, click Run sandbox anchor, and your browser will SHA-256 the synthetic inputs, POST them to the public sandbox API, and render the receipt that comes back — block hash, chain sequence, verify URL, everything. No bearer token, no API key, no procurement conversation. Sandbox receipts carry the explicit chain: "trustai-sandbox" marker inside the cryptographic commitment, so no examiner could mistake one for a production decision receipt.

Public, no API key 100 anchors / hour per source IP Isolated sandbox chain 30-day rotation Same schema as production SHA-256 in your browser

Scenario 1 of 3

A contract clause review.

Your AI agent reviews an indemnification clause and returns a risk score. Your firm wants a tamper-evident record of the model version, the input hash, the output hash, and the reasoning trace hash — in case a discovery request or malpractice query lands eighteen months from now. Click below to anchor a synthetic version of that event.

legal_contract_clause_reviewed

Sample matter · Acme Corp Master Services Agreement

Input text (synthetic) Acme Corp MSA clause 12.3 (indemnification): "Vendor shall indemnify Customer against any third-party claims arising from Vendor's gross negligence or willful misconduct, capped at the fees paid in the preceding twelve months."

Output text (synthetic) Indemnification cap (12 months of fees) is below market for a service of this scope; recommend uncapped indemnity for IP infringement and 24x fee cap for gross negligence. Severity: medium.

Scenario 2 of 3

A proposed redline.

Your AI agent proposes a redline against the same clause. Your firm needs to record which model produced the redline, the SHA-256 of the input, and the SHA-256 of the diff itself — so a year later the firm can prove the redline came from a specific model run, not a later edit.

legal_redline_proposed

Sample matter · Big Tech Inc vendor agreement amendment

Input text (synthetic) Big Tech Inc vendor agreement, section 8.4 (limitation of liability): existing text proposes a $50,000 aggregate cap on direct damages, with all consequential damages waived.

Redline diff (synthetic) Replace "$50,000 aggregate cap" with "the greater of $500,000 or fees paid in the preceding 24 months"; remove the consequential-damages waiver for IP infringement and data breach categories only.

Scenario 3 of 3

A multi-vendor cross-validation divergence.

Your AI agent runs three models in parallel against a regulatory question. Claude says one thing, ChatGPT says something different, Grok says something different again. The substrate anchors the divergence event so your firm has a tamper-evident record that the consensus broke and which indices diverged.

legal_cross_validation_completed

Sample matter · Sample LLC GDPR compliance memo

Decision context (synthetic) Sample LLC GDPR Article 6(1)(f) legitimate-interest balancing test for behavioral analytics processing. Three models run in parallel; consensus status is divergence at positions 2 and 5 of the multi-step balancing test.

What you just saw

The same substrate, the production receipt is just authenticated.

The sandbox API is the public no-auth twin of POST /api/trustai/v1/decisions/anchor. Same payload schema, same hash-chain shape, same verify endpoint pattern — the production surface adds bearer authentication, per-tenant API key scoping, rate-limit tiers, idempotency keys, and ML-DSA-87 post-quantum signatures on every block. The production chain is also anchored to Bitcoin via OpenTimestamps and rotates never.

A successful sandbox anchor is your integration test — if your code can hash the inputs and POST them here, the same code will work against the production substrate as soon as your firm holds an API key.

Read the integration page Request a production key

Try the TrustAI substrate live.Sixty seconds, no procurement.