AI vendors
Your buyers stopped taking your word for it.
Enterprise security questionnaires grew an AI section, and hospital committees now ask for validation evidence. Spectralgraph measures the model behind your product and hands you a signed evidence pack you can put in front of the people deciding whether to buy.
$12,500 PUBLISHED · FIVE BUSINESS DAYS · YOUR REPORT, YOUR CONTROL · FEES NEVER CONTINGENT ON RESULTS
The problem
Deals stall in security review because self-reported evals stopped counting.
The AI section asks for model provenance, hallucination controls, and evidence that outputs get evaluated. An internal dashboard doesn't answer that, because your own team produced it.
The Vendor Evidence Pack
One measurement, reused across every deal for a year.
Evidence report
A signed, dated screening measurement of fabrication behavior in the model behind your product, on your real question domain: where it fabricates, how often, and on which topics. Flagged answers are quoted so your engineers can check each one, with methods, error rates, and limits stated plainly.
Summary letter
A short signed letter for your buyers' reviewers, insurers, and auditors. Yours to hand out. The lab never publishes it and never names you without written permission.
Evidence brief
One page formatted for questionnaire responses and trust-center posting, so your sales team isn't improvising what the letter says.
Written reviewer support
For twelve months your buyers' reviewers can put technical questions to the lab directly and get answers in writing, on the record. The lab describes the measurement and never advocates for a purchase.
The measurement reads the model's own internal signals, so it needs a stack that exposes token probabilities. Its published characterization is 0.852 AUC (fabricated-entity discrimination, length-controlled, leave-one-model-out) in the lab's method paper (DOI: 10.5281/zenodo.21365654).
Reviewers will probe one thing in particular: the measurement can't be coached. Context injection changes what a model says, not the weight-level geometry the score reads, and those experiments are printed in the method paper with their boundary conditions.
Also in the line
Keeping the evidence current.
The Evidence Refresh, $5,000 per quarter keeps it current. Full numbers on the pricing page.
Terms
Your first engagement is invoiced on delivery.
The invoice for a vendor's first engagement goes out with the delivered pack and carries a written guarantee: if the pack gives your sales file nothing you can put in front of a buyer's reviewer, say in writing what is missing within fourteen days and the invoice is cancelled.
Straight answers
The five questions vendors ask first
Will our buyer's security team accept a report from a lab they haven't heard of?
Judge it on its contents: method, controls, error rates, limits, and the exact flagged answers, all reproducible. For twelve months your buyers' reviewers can send questions straight to the lab and get written answers on the record, and a one-page lab brief ships with every pack.
What if the results are bad?
Then you know before anyone else does. The report is yours and stays confidential: the lab never publishes it and never names a client without written permission. You can fix what the measurement found and re-measure within 60 days for a new, separately dated letter.
We run on OpenAI or Anthropic APIs. Can you measure us?
Only where the deployment exposes prompt-side token probabilities, as vLLM-class stacks do. The instrument reads the model's own signals and never grades its text, so a closed API that exposes none leaves nothing to measure. Azure OpenAI exposes probabilities on output tokens only, which isn't enough. Send the model and what serves it, and you'll get a written answer on which case you're in.
Our engineers already run evals. Why pay for this?
Keep them, they're good tools. They just aren't independent, and enterprise reviewers say so out loud now.
Does the measurement need our customers' data?
Never. Probe sets plus a synthetic or de-identified question domain you supply, against an endpoint you designate. Nothing is installed, and everything you share is held under a written agreement.
The lab only needs two facts to start.
Tell us what your product is and who is asking you for evidence. You get a written plan, the engagement letter, and the sample report back the same day.
contact@spectralgraph.ai
REPLIES SAME DAY · EVERYTHING ANSWERED IN WRITING, ON THE RECORD