Tell us which route applies and we will reply on that basis. Which entity an enquiry belongs to is decided in the first conversation, because it determines what we are then permitted to do.
Open to anyone: a regulator, a journalist, an agency relying on a system, an organisation whose members are affected by one, or an operator willing to offer access. Tell us what the system does, who it decides about, and what evidence exists that it is failing someone.
No fee is involved and none can be. Proposing a study does not buy one, and the operator of a system cannot fund its evaluation under any arrangement.
Use this for right-of-reply correspondence, factual corrections, and to tell us about a remediation. A developer who fixes a documented failure is the fastest route to reduced harm we have, and we will publish that response alongside the finding.
Also the route for contesting a published finding after the reply window has closed. Contested findings are reassessed by someone who did not run the study, and the outcome is published whether or not it favours us.
Handled by Quantile Assurance, which is a separate entity with separate accounts and a separate board.
Engaging Assurance makes your systems permanently ineligible as research evaluation subjects. That is not a cooling-off period and it does not expire. It is worth knowing before the first meeting rather than after.
We will talk about anything we have published, including its limitations. We do not comment on studies in progress, on subjects we have not published on, or on systems we have not tested.
We respond to consultations, and we will discuss what an evaluation regime can and cannot establish before it is written into a rule. There is no charge for this and it creates no commercial relationship.
This form is the only route in. Everything is read by a person.
Whether the system is deployed and making consequential decisions; whether ground truth exists or can be built; what access is obtainable and therefore what an evaluation could establish; whether any conflict disqualifies us; and whether the finding would be publishable whatever it concluded. If the answers rule out an honest evaluation, we say so rather than scoping a smaller piece of work around them.