← TrustAI

For legal aid & access to justiceThe institution licenses itWorks on any vendor’s AIInformation, not advice

Say yes to AI — without risking a sanction.

Your program is stretched, and AI is how you do more with the capacity you have. But courts are sanctioning lawyers for AI-hallucinated citations — and a legal-aid client is the last person who can absorb a defective filing. TrustAI is the verification layer that sits between your AI tools and the door: it checks what the AI produced — from any vendor — before anyone relies on it. The institution licenses it; the people you serve never pay.

The risk you’re managing. In Mata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023) a court sanctioned attorneys $5,000 over six AI-fabricated opinions, and the duty to verify AI output is now explicit (ABA Formal Opinion 512, 2024). A guardrail that catches a fabricated citation before filing is what lets a capacity-constrained program adopt AI without taking on that exposure.

What it checks — on whatever your team already uses

Live

Every citation, verified

Each case citation is checked against the CourtListener / Free Law Project database and returned as verified, not found, or unverifiable — and when coverage is uncertain it abstains rather than guesses. It never accuses a real case of being fake, and never invents a verdict.

In the full analysis

Cross-checked, not single-sourced

In the deeper analysis behind a pilot, the work is run past multiple independent models, so a single vendor’s hallucination doesn’t pass silently. Where they diverge, the disagreement is surfaced for a person to judge — not hidden.

Live

Structural review flags

A quotation tied to no citation, or a legal proposition stated with none, is flagged for your review — a structural observation, never a determination that the statement is wrong.

Live

A tamper-evident receipt

Every check produces a receipt that shows the evidence behind each verdict and seals the method. Keep it on file to show how the AI output was checked, if a client or a court ever asks.

It surfaces information; your attorneys judge fit and correctness. It never tells you an answer is right, never opines on whether an authority supports your argument, and never replaces a licensed attorney’s judgment.

Built for how legal aid actually buys

The institution licenses it — the client never pays.

Sources & authority

What this relies onAuthority
The fabricated-citation risk this addressesMata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023)
The duty to verify AI outputABA Formal Opinion 512 (2024)
The citation databaseCourtListener / Free Law Project, Citation Lookup API
The tamper-evident recordKnox chain — hash-linked, independently re-checkable

See it on your own work

The fastest way to judge it is to run one of your real draft types through it and watch what it catches — and where it honestly abstains.

Request a pilot →

Prefer to try the public tool first? The free independent verifier checks citations and flags structural gaps on any AI output instantly — the multi-model cross-check runs in the deeper pilot analysis.

TrustAI is not a law firm and does not provide legal advice. It independently verifies whether citations exist, cross-checks AI output, and flags structural gaps for your review; it does not tell you whether an authority supports your argument, whether the output is correct, or what action to take. Those judgments belong to a licensed attorney. What TrustAI does, and never does →