The OGEMEX Desk
How we use artificial intelligence
We use AI-assisted drafting. A named person is editorially responsible for every published word. Nothing reaches a customer on the strength of a machine output alone: every legal proposition is checked against the primary instrument, every figure is sourced, every correction is published permanently and never removed. We tell you this because a firm selling compliance intelligence cannot reasonably expect its own governance to go undisclosed.
1. Why this page exists
Article 50 of the EU AI Act took effect on 2 August 2026 and requires that AI-generated content be marked and, in defined circumstances, disclosed. Several major model providers now embed machine-readable marks in generated text and signed provenance metadata in generated files, applied globally rather than only within the Union. That is a welcome development and we have no interest in working around it.
It follows that a reader of our material may in due course be able to establish that AI was used in its preparation. We would rather state the position ourselves, precisely, than leave it to be inferred.
2. What AI is used for
Drafting, structuring, translation between formats, code for our own tools, and first-pass research retrieval. It accelerates production. It does not decide what is true.
3. What AI is never used for
- Determining whether a legal proposition is correct. That is established against the instrument itself.
- Generating a citation, an article number, a date or a figure that is then published unverified.
- Deciding what is published. A named human approves every external release.
- Advising on your specific facts. We do not offer legal, tax, audit, accounting, financial or investment advice, and no AI output is presented as such.
4. The editorial controls that sit over it
| Control | What it does |
|---|---|
| Primary-source verification | Every legal proposition, date, threshold and penalty basis is checked against the regulation, statute or regulator publication itself. Where a pinpoint reference has not been read at source, the entry says so rather than carrying an invented one. |
| Confidence marking | Every obligation in OGEMEX Intel carries a confidence level and, where applicable, a flag stating whether the uncertainty arises from an instrument in motion, from national implementation, or from a test that depends on your own facts. |
| Named approval | No customer-facing file is released without an explicit approval step recorded against a named person. |
| Permanent corrections log | Every correction to a previously published position is recorded with what was said, what it now says and why. Entries are never removed. A record that edits its own history is not a record. |
| Pre-publication checks | Automated integrity checks run before release, covering source links, citation coverage, score reproducibility, internal consistency and staleness. A failure blocks publication. |
5. What you are paying for
Not the drafting. The drafting is the cheapest part of this work and always has been. You are paying for the selection of what matters, the verification against source, the assignment of an accountable function, the assessment of consequence, and a named person who stands behind the result and corrects it in public when it is wrong.
A regulatory register that no one has checked is worth nothing regardless of who or what wrote it. That is the distinction this page exists to draw.
6. Machine-readable marks in our materials
Text prepared with AI assistance may carry an imperceptible watermark applied by the model provider. Files may carry signed provenance metadata. We do not remove, strip or attempt to defeat these marks, and we would regard doing so as inconsistent with the compliance standards we ask our own readers to meet.
Such a mark indicates that a model processed the text. It does not indicate the absence of human authorship, editorial judgement or verification, and it says nothing about accuracy. Our editorial record, not the presence or absence of a mark, is the basis on which this work should be judged.
7. Your data
We do not put customer data, client documents or subscriber information into third-party AI systems as part of producing our published materials. The OGEMEX Intel dashboard performs all assessment in your own browser and transmits nothing. See the privacy notice.
8. Challenge and correction
If you believe anything we have published is wrong, write to partner@landmarkglobalpartners.uk quoting the obligation identifier or document title. Corrections are made at the next compilation and recorded permanently in the published corrections log.