Deadline Day — and the Section 31 Wall Appears
Today was supposed to be the day the picture started becoming clearer.
Most of the facial-recognition FOI requests in the Seen. Not Processed. project were submitted on 21 August 2026. Allowing for weekends and the August bank holiday, 21 September was therefore the 20-working-day deadline for the bulk of the police forces we approached.
We were expecting answers.
Instead, a wall appeared.
By the end of the day, 27 police bodies across 26 tracked FOI cases had invoked, considered or flagged Section 31 — Law Enforcement in relation to our request. In many cases, forces extended the response period while carrying out a public-interest test rather than providing a substantive response.
It is important to be precise about what that means. These are mostly procedural extensions, not final refusals. We do not yet know what information the forces will ultimately disclose or withhold.
But the pattern is striking.
Our request was deliberately written to avoid operationally sensitive material. We did not ask for watchlists, identities, covert tactics, passwords, security credentials, exploitable vulnerabilities or network architecture. We expressly accepted appropriate redaction of genuinely sensitive contractual and security information.
The central question was much simpler:
What recorded evidence exists showing that facial-recognition data is actually deleted, retained and audited in accordance with the safeguards the public is told are in place?
During the day, we began challenging vague Section 31 extension notices. We asked forces to identify four basic things: which subsection they were relying upon, which parts of our request were affected, what law-enforcement interest was said to be at risk, and the broad causal connection between disclosure and that risk.
The responses have already been revealing.
South Wales, Cambridgeshire, Kent and Warwickshire all identified Section 31(1)(a) and (b) and specifically pointed to Q5(a), Q5(b) and Q6. Northamptonshire also identified the same statutory limbs and referred more broadly to Q5 and Q6. Several forces used virtually the same explanation: disclosure could cause “prejudice to tactical capability”, making the tool less effective in preventing and detecting crime and apprehending offenders.
British Transport Police provided only part of the clarification. West Yorkshire supplied almost everything except the precise numbered questions affected, so we asked them for that missing piece.
We also spent part of the day correcting dates, narrowing requests where necessary, and updating the tracker so that every procedural step is preserved.
It was frustrating. Instead of simply examining evidence that safeguards worked, we found ourselves repeatedly asking authorities to explain the basis on which scrutiny of those safeguards might itself prejudice law enforcement.
But that process is now part of the evidence too.
Every extension, every clarification, every correction and every eventual disclosure or refusal will be documented.
Because if public confidence in facial recognition depends upon safeguards, the public should be able to see evidence that those safeguards actually operate.
Trust requires evidence.