AIR Certified is the public trust standard for AI tools, the bar a tool must clear before it goes anywhere near someone's real work and data. Assessed against published criteria, decided by an independent council. Any AI tool, any industry. Independent agencies are the first cohort.
Most "certified vendor" programs are directories you pay to appear in. AIR Certified inverts that: you pay for scrutiny, like a product pays a safety lab. The fee covers the assessment labor. It does not, and cannot, buy the outcome.
A flat, public fee priced to the labor of assessing you, not to your company's size or revenue. The invoice says "assessment." Same fee whether you pass or fail, and no refund on a fail. You are buying scrutiny, the way a product buys a UL test.
There is no upsell that turns a fail into a pass, no remediation retainer that quietly grants the badge. Fail, fix the real gaps, and re-apply as new work for a new fee. The standard does not negotiate.
The AIR Standards Council owns the criteria and holds final authority over every pass, fail, and revocation. It holds no equity in Nfinit Monkeys and depends on no certification revenue. The firm cannot grant a pass or block a revocation.
You will never pay a listing or license fee to stay on the registry. When a certification expires, it triggers a fresh assessment, new work against the current standard, not a renewal invoice. The bar moves with the field.
Individual failures stay private, so vendors will actually apply. But every cohort's result is published in the open: assessed N tools, M met the bar. A standard nobody fails is theater. Ours has a published failure rate on purpose.
The standard asks the questions a serious buyer would ask before letting an AI tool touch real work, client data, and reputation. Each dimension has explicit, vendor-blind pass criteria in the public AIR-VS document.
Where your data goes, who can see it, how long it's kept, and whether it trains anyone else's model. Stated and verifiable.
Who owns the outputs, what the model was trained on, and whether using it puts your IP or your clients' IP at risk.
Access controls, encryption, tenancy isolation, incident history, and the posture you'd expect of anything inside your perimeter.
What the system actually does versus what the marketing says. Model disclosure, limitations, and when a human is really in the loop.
How cleanly it fits the stacks you already run, with documented APIs, predictable behavior, and no hidden lock-in at the seams.
Can you get your data and your work out, in usable form, the day you decide to leave? Exit terms in plain sight, not buried.
Every performance and accuracy claim tested against evidence. If the demo can't be reproduced, it doesn't count toward the badge.
Where control stays human: review gates, override paths, escalation, and accountability when the model is wrong.
Uptime, versioning discipline, change notice, and a real support path — so the tool you certify today behaves the same tomorrow.
A single fee, a fixed scope, and a verdict that isn't ours to give. Most programs list you on payment. This one assesses you, then the Council rules.
Submit the tool, its use case, and the surfaces it touches. We confirm fit and lock a fixed, public fee. No size-based pricing.
Flat fee, agreed up frontYou hand over the documentation, configs, and access the AIR-VS criteria call for, across all nine dimensions.
Vendor-blind criteriaAssessors test claims against evidence and score each dimension to the published bar. This is the labor your fee paid for.
You paid for scrutinyThe independent AIR Standards Council reviews the assessment and rules: met the bar, or not. The firm holds no vote.
Pass / fail not for salePasses go on the public registry; the cohort's aggregate result is published. At expiry, a fresh assessment, not a renewal.
Registry is the source of truthBoth tiers meet every dimension. The difference is depth of evidence and the surfaces in scope, not how much you paid. Tiering reflects the assessment, never the invoice.
If you're choosing AI tools, you don't need to take a vendor's word, or ours. Check the public registry yourself, then put the bar in your contract. Drop this clause into any RFP or vendor questionnaire.
The Nfinit Monkeys org-side advisory recommends no specific tools, certified or not. Certification status lives only on the registry you consult independently.
A certification that everyone passes signals nothing. So we hold the line two ways: individual failures stay private, so vendors are willing to apply, and the aggregate is public, every cohort, so the badge keeps its meaning.
Comparable standards reject a large share of applicants. We expect to as well, and we'd rather lose the assessment fee than dilute the bar. The standard never bends.
The criteria are public, so the badge can't hide behind them. Read the AIR-VS standard, then apply to be assessed against it.