Skip to content

Editorial methodology

How our automated Insights briefings are produced

Insights briefings are produced by an automated pipeline and published without human approval. This page explains exactly how that works, and what it does and does not guarantee.

What runs, and when

Once a week, on Monday morning UK time, the engine reads the official feeds listed below, screens what it finds for relevance to UK digital clinical safety and healthcare AI, discards anything already covered or out of date, and attempts a single briefing on the most significant remaining development. It publishes at most one article per week, and frequently publishes none.

Approved sources

Factual claims may only be supported by these primary and official UK sources. Search engines are never queried and copyrighted news reporting is never republished.

  • NHS England and NHS England digital (digital.nhs.uk, england.nhs.uk)
  • GOV.UK publisher pages for the MHRA, DHSC and DSIT
  • NICE (nice.org.uk)
  • Information Commissioner's Office (ico.org.uk)
  • HSSIB (hssib.org.uk)
  • legislation.gov.uk
  • Find Case Law at The National Archives, and the Courts and Tribunals Judiciary

BAILII may be used to notice a judgment, but a judgment is never cited from BAILII alone when the official version is available.

Checks that must pass before anything publishes

  • Every cited link is on the approved list, was fetched successfully during the run, and has a recorded title and publication date.
  • The regulatory status is explicitly classified — law in force, standard, guidance, consultation, proposal, enforcement action, allegation or final judgment — and is stated on the page.
  • No long verbatim quotation is carried over, so a quotation cannot be fabricated.
  • Court coverage publishes only from a complete published judgment or official decision, and must state the court, the decision date and the outcome.
  • Nothing that gives personalised clinical or legal advice, overstates what this practice does, or implies NHS or regulatory endorsement.
  • A deterministic on-page quality score of at least 90 out of 100. This is our own internal measure of structure, sourcing and readability — not a claim about search rankings.

If any check fails, nothing is published and the week is recorded as skipped. The engine is built to publish nothing rather than publish something thin or unsafe.

Limitations you should assume

  • No person reads a briefing before it appears. There is no clinical, pharmacist or legal review.
  • Coverage is not exhaustive. The engine monitors a fixed list of official feeds once a week, so a development outside those feeds will not appear.
  • A briefing is general information about a published development. It is not advice about your product, your organisation or your legal position.
  • Interpretation can be wrong. Always read the primary source, which is linked in full on every briefing.
  • Nothing here implies endorsement by the NHS, the MHRA or any other body.

Authorship

Briefings are published as the Assurance AI Editorial Desk. They are not written by, and are never attributed to, Khalid Anwar or any other named person. Our guides and downloadable resources are a separate thing entirely: they are drafted, reviewed and approved by a person before they appear, and that process is unchanged.