Sector and jurisdiction specific, anonymised, and read against companies
under the same regulatory obligations you answer to.
The kn0w peer dataset is the anonymised aggregate of every audit kn0w has issued. It is the second artefact of every commission — the first being the signed Opinion delivered to the member, the second being the contribution to the dataset the next member is read against.
A consultancy that holds peer AI governance data sells advisory against it. A vendor that holds it is the party whose controls are being measured. A law firm that holds it bills hours against its own data. kn0w does none of these. The dataset is held by an issuer with no advisory retainer, no platform licensing, and no billable-hours line.
The k=5 floor is enforced in the anonymisation pipeline itself, from the first audit commissioned.
Every contribution is made under a single Data Contribution Agreement naming KN0W PTE. LTD. as data controller. No public scraping. No regulatory-filing inference. No simulated peers. The dataset exists because members commissioned audits and consented to their anonymised contribution joining it.
The dataset is cut on three axes: sector, jurisdiction, and staff band. A member is read against the cohort they fall into, not against an average of the whole.
Each cohort operates under the framework its members answer to. An Australian FinTech under APRA and ASIC is not read against a UK HealthTech under the MHRA — the regulatory context, operational shape, and governance obligations differ, and the reading would not be defensible across them. Each cohort is a distinct peer set.
Funding stage is captured at intake but is not a cohort axis. It is self-reported, blurs at the edges, and does not hold across funding events. Sector, jurisdiction, and staff band define the peer set.
Where a cohort holds fewer than five members at the time a reading is produced, that cohort returns insufficient peer data in your cohort rather than a percentile. No partial benchmark is issued below k=5, in any cohort, at any time.
From kn0w's published sample Opinion. Subject company and figures are fictional.
Each new audit contributes to the cohort it sits in, at the k=5 floor, from the quarter it is issued.
A member’s reading is computed against the dataset as it stood at issuance. At the institutional anniversary, the Annual Statement re-reads the member against the dataset as it then stands — issued under the same hallmark, signature, and methodology as the Opinion. The dataset a member is benchmarked against is the dataset on the day the artefact issued, not a figure that moves under them between issuances.
The construct behind the reading — the six dimensions, their weighting, the k-anonymity floor, the bias controls, and the regulatory frameworks mapped into the instrument — is documented in full at /methodology.
Data handling and residency are documented at /security. The dataset is governed under a single Data Contribution Agreement naming KN0W PTE. LTD. (Singapore) as data controller.
The benchmark exists because members commissioned audits and consented to their anonymised contributions joining the dataset. It grows with every audit issued. No member is read against an industry average, a synthetic composite, or a vendor-held figure. They are read against their peers, under the framework their regulator names them accountable for.