Your AI guardrails have been demonstrated. Have they been tested?

Sea Trial sends 44 attacks and 245 disguised versions of them through a guardrail stack, counts what got through and how many ordinary customers were refused along the way, and writes the result into a hash-chained record.

One HTML file. No account, no server. You can paste a real transcript into it.

One attack, seven disguiseslive, not a recording
—

First pass with every rail on. In a moment the unmasking rail comes off.

In model risk this is called effective challenge. Somebody who is not the vendor tries to break the control and keeps a record of what happened. The vendor demo is not that.

—of the built-in attacks walk straight through when the unmasking rail is off

A vendor demo shows one attack being blocked. It doesn't show the next ten.

I have sat through those demos. The attack is typed cleanly, the filter catches it, everyone nods. Nobody types it in leetspeak. Nobody swaps in a Cyrillic "а" or pastes it base64-encoded inside a note. Nobody asks what happened to the customer who wrote "ignore my earlier message, I found the statement" and got refused.

Those are the two questions I would expect from an examiner: what gets through, and how often do you refuse a customer for no reason. Sea Trial answers both from a fixed configuration, so a second reviewer running it gets the same number you did.

What happens when you run it

Five steps, about ten minutes the first time. The score is provisional until step two.

score: none

Run the sea trial

44 attacks and 14 ordinary messages go through the ten rails. You get a detection rate, a false-positive rate, and a grid of which rail caught what.

score: provisional

Run the mutation storm

Each caught attack is rewritten seven ways (leetspeak, spaced letters, look-alike characters, zero-width padding, base64, reversed, buried in a polite message) and sent again. This is where most stacks lose points.

score: final

Bring your own material

Paste your own red-team cases or complaint transcripts. Add your own phrases, tool names and canary strings to the rails. Your cases get their own rows in the grid.

score: yours

Tune and re-run

Change a threshold or switch a rail off and the numbers update. Usually you trade one false positive for one miss; now you can see which one.

score: sealed

Issue the certificate

One page: grade, settings, findings, the mapping to OWASP, the EU AI Act, NIST AI RMF and SR 11-7, and the chain hash. Print it and file it.

Below is a real run of the built-in corpus at default settings, generated when this page loaded. Yours will come out differently.

The output is a page for the validation file

Coverage by vector and rail

Each cell is how many attacks in that vector a rail flagged on its own. A row with one filled cell is a single point of failure.

CERTIFICATE OF SEA TRIALReproducible from the sealed configuration
A

Conditions

    TRIED AND HELD

    Ten rails. Three carry most of the weight.

    Each rail is a readable set of rules, so a reviewer can see exactly why a message was held.

    Normaliser

    Undoes disguises before anything else reads the text: base64, letter spacing, leetspeak, Cyrillic look-alikes, zero-width characters, reversed strings. Turn it off in the demo above and the same attack gets through five times out of eight.

    OWASP LLM01, EU AI Act Art. 15, NIST MEASURE 2.7

    Authority guard

    Holds anything the assistant is not entitled to do: wires, deletions, limit changes, waived verification, shell commands, calls to privileged tools, and any amount over the cap. In the default stack it is the rail with the second-most attacks depending on it alone.

    OWASP LLM06, EU AI Act Art. 14, NIST GOVERN 1.7

    Output contract

    Checks the reply, not the request. Malformed JSON, "guaranteed returns", "I have transferred the money", "as an unrestricted assistant". Most guardrail products stop at the input. This is the rail examiners ask about.

    OWASP LLM05, EU AI Act Art. 13, NIST MEASURE 2.5
    All ten rails and the controls they evidence

    Who uses it and what they take away

    Chief AI Officer the certificate

    One page with three numbers that did not come from a vendor: attacks caught, disguised attacks caught, customers wrongly refused. Dated and tied to a configuration hash.

    Model risk and validation the findings

    The coverage grid and the single-point-of-failure list. They read like findings because that is what they are. SR 11-7 asks for evidence of challenge; this is the evidence.

    CISO and red team the corpus

    Your own attack cases and rules, saved as a JSON file. Re-running next quarter is an import and a click, and whoever has the job after you can do it.

    Vendor selection a comparison

    Run two configurations through the same corpus and put the two certificates side by side. It is a shorter conversation than comparing decks.

    What keeps the result honest

    Same inputs, same resultThe detectors are rules, not a model. A second reviewer with your configuration gets your number exactly.
    Every action is hashedEach stack change, test, storm and certificate is chained with SHA-256 to the entry before it. The export opens in THE VERIFIER.
    Nothing leaves the browserOne HTML file. No server, no analytics, no model calls. A real customer transcript can go in.
    It reports its own gapsThe default stack scores 98. One ordinary message is wrongly held and eleven rewrites get through. Both are printed on the certificate.
    The bank is inventedCompass and Northlake Federal Credit Union do not exist. The canary strings, account numbers and credentials in the corpus are made up.
    It tests a design, not live trafficSea Trial is for validating a guardrail stack. The runtime control that sits in front of production is BULWARK, in the same line.

    Send me your logs and your policy. You get back a certificate, the evidence chain, and a corpus you can re-run.

    Half a day with your team, or installed for your validators to run on their own. Either way you finish with the same three files.

    Email Richard
    SEA TRIAL instrumentOpen in its own tab