Skip to content
kaniyan.ai

Trust

How we use AI

We ask institutions to disclose when a machine took part in a decision. This is us doing the same, for our own website and our own work.

Effective
27 August 2026
Version
1.0

On this website

Written entries we publish are produced in one of three ways, and every entry says which, on the page, above the fold of its footer:

  • Written by a person. No agent involved.
  • Drafted by an agent, edited by a person. An agent produced the draft; a named person edited and approved it and is accountable for every claim in it.
  • Written by an agent. Reserved for mechanical content — changelogs, generated summaries. Still reviewed before publishing.

Where an agent was involved, the page names the charter and version that ran, and the person who reviewed it. That is not a nicety: the database refuses to publish agent-drafted content without a named human reviewer and a timestamp. The control is a constraint, not a checklist item.

What a person always does

  • Decides what is worth writing about.
  • Verifies every factual and numeric claim against its source.
  • Approves publication, by name.
  • Owns the consequences if it is wrong.

No agent publishes to this site. Nothing reaches a reader without a person having read it first.

What we will not do

  • Fabricate a reference. If a claim has a source, the source is linked. If it does not, the claim is marked as our opinion.
  • Generate a case study. Case studies describe work we actually did, with the customer’s agreement.
  • Use a synthetic image of a person or an invented quote, in any context.
  • Pass off a statistic we have not verified. Numbers in our writing come with a source or come with the word “roughly”.

In our client work

The same rule applies at a larger scale, and it is the product: an agent proposes, a named officer approves, and the record shows what the agent saw, what it concluded, why, and who signed. We do not deploy agents that make consequential determinations about people without a human gate, and we will decline work that asks us to.

Why publish this

Because a company that sells accountability infrastructure and does not disclose its own use of AI has a credibility problem it cannot argue its way out of. And because writing it down is the only way to be held to it.