TRUSTED BY LAWYERS, STUDENTS, AND COMPLIANCE TEAMS

The problem isn’t that AI is wrong.
It’s that we believe it’s right.

46 verified incidents. Billions in damages. Most were preventable with verification.

Court Cases

$0B

MN Fraud Alone

Worst AI Error Rate

ZH-1 Error Rate

ZH-1 catches fabricated citations, verifies regulatory eligibility, and scans documents for hallucinated facts — so you never file, submit, or publish something that isn’t real.

Lawyers: Verify every citationStudents: Check every sourceGov agencies: Compute eligibility from lawGrant writers: Find what you qualify for

Try the Live Demo See How It Works

Verified Checks

Output Variance

Hallucinations Found

Results are computed from encoded rules, not AI generation. External data sources are subject to standard validation.

zh verify

$ zh verify --rule "7 CFR 273.9"

[1/4] Loading ruleset... done

[2/4] Evaluating input... done

[3/4] Computing decision... ELIGIBLE

[4/4] Hashing audit record... SHA-256 chained

Result: ELIGIBLE

Basis: 7 CFR 273.9(a)(1)

Variance: 0.0% | Hash: a3f8...c2d1

WHY THIS MATTERS

This is what happens when you trust AI without verification.

Lawyers sanctioned. Patients endangered. Businesses fined. These are real stories — and they happened because nobody checked whether the AI’s answer was actually true.

Stories sourced from public news coverage. Cost descriptions are editorial characterizations, not direct quotes.

FBI / DOJ2025

Minnesota: $250M Stolen via Fake Meal Rosters

Cost: $250M stolen, 79 charged

CNN / CNBC2025

Character.AI Linked to Teen Suicides

Cost: Settled (undisclosed)

Bloomberg2025

Anthropic Settles for ~$1.5B Over Pirated Training Data

Cost: ~$1.5B settlement

VinciWorks2025

Largest AI Court Sanction: $85,841 Fine

Cost: $85,841 sanction

Columbia / Tow Center2025

Grok 3 Got 94% of News Citations Wrong

Cost: 94% citation error

CO Legislature2026

Colorado AI Act Takes Effect

Cost: Compliance required

DOCUMENTED FAILURES

46 verified AI failures. Real costs. Real consequences.

From $85K court sanctions to $250M fraud schemes — we documented every case so you don’t become the next one.

Legal2025

ByoPlanet v. Johansson

$85,841

Government2025

Minnesota: Feeding Our Future ($250M+)

$250M+ Stolen (79 Charged)

Healthcare2025

Character.AI Teen Suicide Lawsuits

Settled (Undisclosed)

Explore the full case database

HOW IT WORKS

We don’t guess. We compute.

Most AI tools generate probable answers. ZH-1 looks up the actual rules and computes the actual answer. Same question, same rules, same result — every single time.

Same input. Same answer. Every time.

Our engine evaluates real regulatory rules — not probabilities. There’s no language model guessing at your answer. Ask the same question twice, get the same result twice. That’s what makes it safe to rely on.

0.0% variance

Every decision has a receipt.

Each result is logged, timestamped, and cryptographically chained to the one before it. If anyone changes a historical record, the chain breaks. You get a tamper-proof paper trail for every check you run.

SHA-256 chained

Your auditor can verify it themselves.

Every decision shows exactly which rules were applied and why. No black boxes, no “the AI said so.” Hand the report to your auditor, your professor, or your regulator — they can independently confirm the result.

Full audit trail

TRY IT NOW

See the difference for yourself.

Run a compliance test to see ZH-1 vs. LLMs, or scan a document for hallucination patterns.

ACCURACY SCOREBOARD

— compliance tests. Live results.

ZH-1 is a deterministic engine — it computes answers from rules. The models below generate answers probabilistically. This demonstrates why deterministic verification exists.

ZH-1 is a rule-based engine, not a language model. Its accuracy reflects currently encoded rulesets and does not guarantee regulatory completeness. AI model scores reflect their responses to the same compliance scenarios. Results update as rulesets are expanded.

ZH-1

0.0%

Hallucinations0

Variance0%

Speed47ms

Claude Opus 4

0.0%

Hallucinations158

Variance34.2%

Speed2340ms

Llama Maverick

0.0%

Hallucinations255

Variance41.7%

Speed3120ms

Gemini 3 Pro

0.0%

Hallucinations204

Variance38.9%

Speed2780ms

Results from deterministic testing of — compliance scenarios. Deterministic outputs reflect currently encoded rules and do not guarantee regulatory completeness. See Terms of Service.

WHO IT’S FOR

Real answers for the questions that matter most.

Housing Compliance

Is this family eligible for Section 8? Should this voucher be renewed? ZH-1 computes HUD eligibility from 43 encoded rules across 5 programs — no guesswork, no judgment calls.

Government Benefits

SNAP, Medicaid, SSI, TANF — each with rules that change every year. ZH-1 stays current so applicants get the right answer the first time.

Grant Eligibility

3.1M

Stop guessing whether you qualify. ZH-1 checks your org against 3.1M federal opportunities on SAM.gov and tells you exactly which grants match.

Academic Citations

150M

Did your AI tool fabricate that source? ZH-1 checks every citation against 150M scholarly records in the CrossRef database so you never submit a paper with fake references.

Medical Compliance

Coming

Healthcare cannot afford 'probably correct.' Rule encoding is in progress — not yet available for PHI or HIPAA-regulated workflows.

Custom Rules

∞

Have your own compliance rules? Encode them into our engine and get the same deterministic, auditable results for your domain.

TRUST & SECURITY

Built for regulated industries.

Here’s exactly what we do and don’t do with your data and our architecture.

Data Security

No document content stored after processing
No training on customer data
Encrypted in transit and at rest
Architecture designed for SOC 2 Type II compliance

Audit Architecture

SHA-256 hash-chained audit trail
Every decision cites specific rules
Tamper detection is mathematical
Independent third-party verifiable

Compliance Ready

Deterministic outputs for auditors
Full processing integrity evidence
Change monitoring built in
Patent pending (US utility application filed)

BY THE NUMBERS

Built in public. Verified in production.

Housing agencies

Benefits offices

Grant writers

Universities

—

compliance checks run

100%

decision accuracy

—

audit records created

The ZH Standard is live and processing real compliance checks. We’re currently onboarding early adopters across housing authorities, benefits agencies, grant compliance teams, and academic institutions.

PRICING

Start free. Scale when you’re ready.

Every plan includes citation verification, hallucination scanning, and regulatory checks.

Free

SCANNED

$0forever

20 checks/mo

No overage

Hallucination scan

Academic citation verification

Hedging language detection

Upload text, PDF, docs

"Scanned by ZH-1" badge

Starter

SCANNED

$29/month

200 checks/mo

$0.10/check after

Everything in Free

Government benefits verification

Grant eligibility checks

Current with 2025 law changes

Priority scanning

Questions we hear most.

GET STARTED

Stop wondering if the AI got it right.

Create a free account and start scanning documents in under a minute. No credit card required.

The problem isn’t that AI is wrong.It’s that we believe it’s right.

This is what happens when you trust AI without verification.

46 verified AI failures. Real costs. Real consequences.

We don’t guess. We compute.

See the difference for yourself.

— compliance tests. Live results.

Real answers for the questions that matter most.

Built for regulated industries.

Built in public. Verified in production.

Start free. Scale when you’re ready.

Questions we hear most.

Stop wondering if the AI got it right.

The problem isn’t that AI is wrong.
It’s that we believe it’s right.