Modulith Lab Cert

Get your AI implementation assessed against explicit formal standards

Connect your model endpoint to the Modulith API. The assessment runs blind — like any laboratory, we do not know whose sample we are testing until after the result is recorded.

Blind assessment · Fixed scope · Fixed price · No meeting required

Designed like a lab test, not a consulting engagement. The protocol is fixed in advance. You do not meet the lab to adjust the test or negotiate the result.

Blind assessment

We do not know whose sample we are testing

When you connect your model endpoint, Modulith does not know who you are. Your implementation connects under a session ID — no name, no company, no identifying information. The 223 queries run. The results are scored automatically using Lean 4 verified logic. The outcome is locked into an Overview ID.

Only after the result is locked do you add your company name and team label. These are display metadata on the final report. They never affect results. There is no mechanism for bias to enter — this is not a policy, it is how the system works.

This is the same principle used in clinical laboratories, academic peer review, and forensic testing: the assessor does not know whose sample is on the bench. The result is the result.

How it works

From connection to Lab Cert in five steps

Connect your model endpoint

Connect your model’s API endpoint to the Modulith testing API. Whether it’s an off-the-shelf model, a fine-tuned variant, or a wrapped application with retrieval and guardrails — if it has an endpoint, it can be assessed. For sensitive or air-gapped implementations, use the manual submission page instead.

Agree to the assessment disclaimer

Confirm that the model endpoint or responses you provide accurately represent your deployed implementation. This is your attestation — if the test is gamed, the report is worthless to you, and the liability sits with you.

Blind assessment runs automatically

Modulith sends 223 queries to your endpoint and scores the responses against all 10 formal standards. The assessment is blind — Modulith does not know who you are. Results are locked into an Overview ID. You receive a free overview showing headline PASS/FAIL results. Run the overview as many times as you need — test new versions, configurations, or wrappers at any time. Each run is free.

Add your identity and purchase

If you want the formal artifact, purchase a Lab Token. Only now — after the results are locked — do you add your company name and team label. Identity fields are display metadata only. They never affect results. This is how Modulith ensures the assessment cannot be biased.

Receive your Lab Cert

The locked Overview ID becomes a full Modulith Lab Cert — with report ID, per-property evidence, scoring rationale, and operational note. A fixed artifact your governance team, procurement officer, auditor, or reviewer can read, compare, and file.

Pricing

Fixed scope, fixed price, same for every customer

Every Lab Cert follows the same methodology, the same queries, and the same scoring logic. The price is fixed and the same for every customer, every implementation, every time. No negotiation. No meeting required.

The free overview is unlimited — run as many times as you need. Test new versions, wrappers, or configurations at any time. When you are ready to issue the formal artifact, purchase a Lab Token.

Get pricing and start your assessment

Enter your email to receive pricing details, vendor license tiers, and instructions for connecting your model endpoint.

No spam · Pricing + setup instructions only

Deliverable

What your governance file receives

The Modulith Lab Cert artifact

A formal assessment report that stands alone without staff explanation. This is the artifact your governance, procurement, or vendor diligence file receives.

Report ID and verification metadata
Per-property PASS/FAIL with measured values
Evidence samples for each finding
Scoring rationale and threshold reference
Methodology version and spec version
Lean 4 verification status
Scope boundary and limitations
Operational note (non-formal observations)

View sample report →

For sensitive implementations

Manual submission for air-gapped or classified environments

If your implementation cannot be connected to an external API — because it is air-gapped, classified, or subject to strict network controls — Modulith provides a secure manual submission page.

You receive each query in sequence, run it against your model on your own infrastructure, and paste the response. Same 223 queries. Same scoring logic. Same blind assessment. No network connection between Modulith and your model at any point.

The assessment follows the same path: results are locked into an Overview ID, identity is added only after locking, and the Lab Cert is issued from the locked result.

Contact lab@modulithlab.com for manual submission access.

For AI model vendors

Vendor furnishing and public reference licenses

If you are a model provider and want to furnish Lab Cert results to your customers or reference them publicly, additional license tiers are available. These allow you to include assessment results in your sales materials, customer documentation, or public communications.

All licenses are based on the same underlying assessment. The methodology does not change. The result does not change. The license determines how the result may be used and referenced.

Contact lab@modulithlab.com for vendor license enquiries.

Continuous monitoring

Monitor API — ongoing safety evidence after certification

A Lab Cert is a point-in-time assessment. Models change. Wrappers get updated. Configurations shift. The Monitor API provides continuous runtime evidence against the same formal standards — so your governance file stays current.

Per-query safety evidence running against your deployed implementation. Tracks behavioural drift over time. Alerts when performance drops below certified thresholds. Creates a timestamped audit trail that regulators and auditors can reference.

The Monitor API does not wrap or modify your model. It observes and measures. Same Lean 4 verified standards, applied continuously instead of once.

Contact lab@modulithlab.com for Monitor API access and pricing.

What a Lab Cert does and does not cover

A Lab Cert tests your implementation's outputs under controlled conditions. It does not audit organisational procedures, deployment processes, governance frameworks, or internal controls.

Some properties are structural. If accuracy, robustness, or bias are not right at the model level, they become very hard to mitigate later with policies or guardrails. Get the foundations assessed independently. Then build your AI safety policy around a result you can trust.

A PASS result means the implementation satisfied a specific standard under specific tested conditions. A FAIL means it did not. Neither is a blanket statement about every possible risk or deployment scenario.