Most AI products do not tell you what the AI did. They produce an output and ask you to trust it. We do not think that posture is honest for a legal product. The LawSensai Trust Center is the public surface where we publish what each AI agent does, how often it does it, what the safety findings have been, and how to verify the behavior. The index lives at lawsens.ai/trust and the per-area pages at /trust/brain, /trust/family, /trust/criminal-defense, and /trust/personal-injury.
This post explains what each page publishes, the audit-log hash chain that makes the numbers verifiable, and how to use the Trust Center if you want to dig into a specific decision.
What the Trust Center does
The Trust Center is the public report on what the AI inside LawSensai is doing. Each per-area page publishes three things. The aggregate decision counts. The safety findings. The runbook for the underlying agent system.
The pages are not marketing. They are a running record of what the agents do. We update the counts on the schedule the page documents. We publish safety findings whether they are positive or negative.
The four per-area Trust Centers
lawsens.ai/trust/brain. The Trust Center for the ANS Core. The Core is the routing engine behind Attorney Match and the Public Defender warm-intro. The page publishes the count of routing decisions, the acceptance rate of warm intros, the PD fallback count, and the safety findings against the per-cohort federated bandit signals.
lawsens.ai/trust/family. The Trust Center for the Family Law product. The page publishes the count of mediation sessions started and completed, calculator runs by state, Settlement Composer drafts produced, guardrail scan flags and their resolution, and Document Vault retrieval counts.
lawsens.ai/trust/criminal-defense. The Trust Center for the Criminal Defense product. The page publishes the count of matters intake, Evidence Checklist usage, PD warm-intro counts and acceptance rates, and Court Date Tracker reminder delivery.
lawsens.ai/trust/personal-injury. The Trust Center for the Personal Injury product. The page publishes the count of demand drafts produced, pi_state_rules version applied at draft time, statute of limitations alerts surfaced, and the safety findings against the treatment summary inputs.
The index page at lawsens.ai/trust links all four and publishes the cross-product aggregate.
The aggregate decision counts
The counts are real. They come from the brain_decisions table that records every AI generation across the product. The brain_decisions row captures the agent that produced the decision, the inputs the decision was based on, the output, and the hash chain entry.
The Trust Center aggregates counts at the level the page publishes. Counts by agent. Counts by practice area. Counts by state where relevant. Counts by safety category.
No individual matter is published. The Trust Center publishes aggregate counts only. Individual matter facts stay private to the user.
The safety findings
The safety findings section publishes what the safety machinery has caught. Prompt injection attempts blocked. Guardrail scan flags issued. Calculator anchors that surfaced a math drift. Routing decisions that fell back to the PD path. Where applicable, we publish the rate at which each finding has occurred.
We also publish failed safety events. If a guardrail scan should have flagged an issue and did not, we publish the post-incident finding and the change we made.
The runbook
Each per-area page includes the runbook for the underlying agent system. The runbook documents the agent's responsibilities, the inputs it operates on, the outputs it produces, the guardrails that constrain it, and the escalation path when something goes wrong.
The runbook is the same document our team uses to operate the system. We publish it because the people relying on the product deserve to see how it works.
The audit-log hash chain
Every decision in brain_decisions is included in the audit-log hash chain. The chain is append-only. Each entry hashes the previous entry so any tampering with a historical decision would invalidate the chain from that point forward. Support can verify the chain at any time.
Decision provenance is queryable by support. If you have a question about a specific decision the product made for your matter, our support team can pull the exact decision record, including the inputs, the agent that produced it, and the chain position. The provenance lookup is how we honor the transparency the Trust Center promises at the aggregate level.
Spanish language support
The Trust Center is published in Spanish at the same URLs with the locale prefix. The aggregate counts, the safety findings, and the runbook are localized. The audit-log hash chain itself is language-neutral.
How to use the Trust Center
Use the index page to see the cross-product picture. Use the per-area pages to dig into a specific product. Read the runbook before you rely on an agent in a high-stakes matter. Check the safety findings section for the most recent updates.
If you have a question about a decision the product made for your matter, contact support and reference the decision. We can pull the provenance from the audit-log hash chain and tell you exactly what the agent saw and did.
What the Trust Center does not do
The Trust Center is not a substitute for using the product carefully. The aggregate counts tell you what the system is doing in general. They do not tell you whether a specific output is right for your matter. Treat AI outputs the way you treat any other draft input to a legal decision. Read them. Edit them. Have an attorney review them when the stakes warrant.
Common misreads we see new users make
Misread one: thinking the Trust Center publishes individual matter data. It does not. The published counts and findings are aggregate. Individual matters are private.
Misread two: assuming a high safety-finding count means the system is broken. The counts include caught events. Prompt injection attempts blocked are a sign the defense is working. Read the rate alongside the count.
Misread three: treating the runbook as marketing. The runbook is the operations document. We publish it because the people relying on the product deserve to see how it actually runs.
Practical next steps
Step one: read the index page at lawsens.ai/trust to see the cross-product picture.
Step two: open the per-area page for the product you use most. lawsens.ai/trust/brain for routing. lawsens.ai/trust/family for Family Law. lawsens.ai/trust/criminal-defense for Criminal Defense. lawsens.ai/trust/personal-injury for Personal Injury.
Step three: if you want to verify a specific decision the product made on your matter, contact support at lawsens.ai/support and we will pull the audit-log entry.
How the Trust Center connects to the rest of LawSensai
The Trust Center is the public report on the entire AI surface. Every product feature, from Attorney Match to the Court Date Tracker to the Settlement Composer to the PI Demand Draft Generator to the Document Vault, writes to brain_decisions. The Trust Center publishes the aggregate. The hash chain makes the record verifiable. The runbook documents the responsibilities. Each per-area Trust Center mirrors the structure of the product area it covers, and the index ties them together.
This post is informational and is not legal advice. The Trust Center publishes what the AI is doing. The decisions you make on your matter, including which outputs to rely on and when to engage an attorney, are yours.
Read more
- lawsens.ai/trust
- lawsens.ai/trust/brain
- lawsens.ai/trust/family
- lawsens.ai/trust/criminal-defense
- lawsens.ai/trust/personal-injury
- National Institute of Standards and Technology AI Risk Management Framework overview at nist.gov
Last verified: 2026-04-09.