The tools component covers data preparation, the categories of ML algorithms, the optimisation methods and the evaluation metrics. Amendment 1 adds transformer and generative model material here.
What the base framework establishes
This is the largest technical component in the framework. The public contents gives it five titled subclauses: general, data preparation, categories of ML algorithms, optimisation methods and evaluation metrics. The algorithm group is where the framework explains the families a practitioner chooses between, and the metrics group is where a result becomes comparable. The public draft of Amendment 1 inserts two new entries in the algorithm group, for transformer architecture and for generative models, and it does not move the existing structure. That placement is the clearest public statement of what the amendment is: an extension of the vocabulary at the point where a practitioner selects a model family.
Why it matters to an AI management system
For a governance lead, this is the clause that turns a model choice into a documented decision. The amendment gives generative models a named place in that decision, which means a description that omits them is now visibly incomplete rather than merely terse.
The public draft adds material at 6.5. Its final wording is not held, so this page records what the draft summary states and does not present it as published text. The public draft summary states that 6.5.3 gains entries 6.5.3.2.9 (transformer architecture and algorithm) and 6.5.3.2.10 (generative model). It does not state any other change in 6.5.
Evidence a reviewer samples
| Category | Artefact |
|---|---|
| Algorithm decision record | The family chosen, the alternatives considered and the reason. |
| Evaluation metrics record | The metrics used, the baseline and the threshold for acceptance. |
| Amendment impact note | Which systems use transformer or generative families and what changes in their description. |
One question to prepare for. Why this model family rather than another, and which metric decided it?
Common gaps
- The algorithm choice is recorded as a preference, with no alternatives and no reason.
- Generative systems are described under a generic model heading, so the new material has no home.
- Metrics are reported without a baseline, so a number cannot be judged.
Prepare before publication
Add an algorithm decision record to the model card and a generative family section to the description. Mark the two new draft entries as the place the amendment lands.
Dates that matter
No separate date attaches to 6.5; the record dates are the stage date, the planning estimate and the CEN expectation stated above.
Cross-references
3.2 Tools · 6.2 Task · 6.3 Model · 8.4 Modelling
Source
Source for 6.5: ISO/IEC 23053:2022, clause 6.5. The public sample stops before 6.5, so this page relies on the public contents and the public summary and does not quote the clause. The two official records are linked on this page so any line can be checked.
Not legal advice; does not certify. A reading of the framework and the evidence a reviewer commonly samples, not a substitute for the standard or your auditor's rules.