NEW RESEARCH: Your Sandbox Is Made of Glass

Read

Trinitite

PricingResearchBlogPodcasts

Free Sample Assurance Audit

See exactly where your AI fails — before an auditor does.

Send us a sample of your real AI chat logs. We score every one against your policy and send back a signed report: the pass rate, the dollar liability, and the exact controls each failure broke. It is free, once a year, and sized to your company.

No card · No contract · Your logs never touch this page

Signed Sample Report

job sa_a1b2c3d4e5 · sampled engagement

Signed

0.0%

Pass rate

Failed logs

0

flagged in this sample

Total liability

$0

summed across failures

Top violated controls

soc2:CC6.6

9

eu_ai_act:art_13

3

hipaa:164.312

2

KMS-signed · Merkle-anchored
third-party verifiable

85%

of AI deployments would fail an audit today

pass / fail

a verdict on every log — not a confidence score

$0

to run it — once a year, no card, no contract

signed

every report is cryptographically verifiable

Why we give it away

A free sample beats a sales pitch.

Most teams have no idea how their AI would score under audit until the audit arrives — and by then a finding is expensive. So we run a real one, on your real logs, for free. You get a signed report whether or not you ever buy a thing. We have found that the report does the selling for us.

Three steps

From signup to signed report.

01

Tell us where to send it

Your name, your email, and the size of your company. That is the entire form. We use the size only to decide how many logs to sample — and we never publish it.

02

Upload a sample — privately

We email you a secure, one-time link on our app subdomain. You drop in a slice of your real AI chat logs there. They never pass through this website.

03

Get a signed report back

A private auditor model scores every log against your policy. You receive a cryptographically signed report you can hand straight to an external auditor.

Bring it in any shape

Any model. Any agent. Any framework.

You should not have to change a thing to find out where you stand. Send the logs you already keep — from any provider, orchestration framework, or home-grown agent. We normalize every format into one canonical shape before the auditor sees a single line.

OpenAI
Anthropic
Bedrock
Vertex AI
LangChain
CrewAI
MCP tools
Custom JSON
+ whatever you log

Format-agnostic ingest

Whatever your AI speaks, we normalize it

OpenAIAnthropicBedrockVertex AILangChainCrewAIMCP toolsAny custom JSONCanonicalnormalizerone shapeAuditorSLM

No re-platforming. No SDK rewrite. Send what you already log.

What lands in your inbox

Not a score. A signed, defensible report.

Every field below comes straight off the audit — the same report our paying customers hand to their auditors.

Pass / fail verdict on every log

Each sampled conversation is scored pass or fail against your stated policy — no "73% confidence," a verdict.

The exact controls each failure broke

Every failure is mapped to the named controls it violated — SOC 2 CC6.6, HIPAA 164.312, EU AI Act Art. 13 — with the clause text quoted.

Dollar liability, summed

Each failure carries an estimated liability figure. The report totals them, so you walk away with a number your CFO can act on.

A signature anyone can verify

The whole report is KMS-signed and Merkle-anchored. A third party can prove it is authentic and untampered without any access to us.

The questions it settles

Walk into the room with answers, not adjectives.

Every audit lands the same way: queued, scored, completed, and signed. What you carry out of it is the part that matters.

QueuedYour sample lands.RunningEvery log scored.CompletedVerdicts + liability.Signed & anchoredYours to defend.

“Where is our AI actually failing?”

The precise conversations that broke policy — and the named control each one violated.

“What is it costing us?”

A summed dollar-liability figure you can take straight into a budget conversation.

“Can we prove any of this?”

A KMS-signed, Merkle-anchored report a third party can verify without trusting us.

“Are we ready for the real audit?”

A rehearsal against your real logs, mapped to SOC 2, HIPAA, and the EU AI Act.

Sized to you

The bigger your footprint, the deeper we sample.

Tell us your revenue band on the form. We scale the sample to match it — and send you the exact details with your upload link. It costs the same either way: nothing.

Growth

Under $100M annual revenue

A focused sample, sized to surface the patterns that repeat across your AI before they become a finding.

Most common

Enterprise

$100M – $1B annual revenue

A serious cross-section — enough volume to quantify your exposure with real confidence.

Global

$1B+ annual revenue

Large-scale sampling that holds up as evidence in front of an auditor or a regulator.

Built for the paranoid

A free audit should not cost you a breach.

This is our public site, so we built the offer to keep anything sensitive far away from it.

Your logs never touch this site

This page collects a name, an email, and a company size — nothing else. Sensitive logs are uploaded later through an authenticated link on a separate subdomain.

Nothing sensitive lives in the browser

There are no backend keys, tokens, or customer identifiers in this page. It cannot reach our control plane, by design.

Evidence you can defend

Each verdict is signed and chained. The report is built for the moment an auditor or regulator asks you to prove it.

One form, one year, one report

Find out what your AI would cost you.

The teams who run this audit stop guessing. They walk into the next board meeting with a signed number instead of a hunch. Start with three fields.

Signed Sample Report

job sa_a1b2c3d4e5 · sampled engagement

Signed

0.0%

Pass rate

Failed logs

0

flagged in this sample

Total liability

$0

summed across failures

Top violated controls

soc2:CC6.6

9

eu_ai_act:art_13

3

hipaa:164.312

2

KMS-signed · Merkle-anchored
third-party verifiable

Claim your free sample audit

Tell us where to send it. No card, no contract — you only give us your logs after we email you a secure, one-time upload link. They never touch this page.

Sets the size of your free sample. We send the exact details after signup.

Your logs never touch this page. We email a secure, authenticated upload link.

Straight answers

Questions, answered plainly.

Is it really free?

Yes — one sampled audit per company, per year, with no card and no contract. We do it because the report sells the platform better than any deck we could write.

How big is the sample?

We size it to your organization. Pick your revenue band on the form and we send the exact sample details — along with your secure upload link — right after you sign up.

Do I need an account on your app?

No. The free audit runs without you logging into the Trinitite app at all. You only ever see the secure upload link and your finished report.

What format do my logs need to be in?

Whatever you already have. Any model, any agent, any framework — OpenAI, Anthropic, Bedrock, Vertex, LangChain, CrewAI, MCP tool calls, or your own custom JSON. We normalize it on our side, so there is nothing for you to rewrite.

Who sees my data?

Your sample is audited by a private auditor model provisioned for your organization. The signed report is yours; you decide who you hand it to.