5.5 Clause
Hybrid AI
The family covers systems that combine methods: learned components beside rules, symbolic reasoning over a model output, or several models coordinated in one pipeline. The draft also defines conversational learning as a hybrid method that improves through dialogue.
What this settles
Hybrid systems are where a single-method description fails most visibly. The interesting behaviour often sits in the handover between components, and so does the accountability question.
Who it binds
The organisation, as a method choice. The clause creates no duty, but it removes the excuse that a system is too mixed to describe.
What to do in your practice
Describe the seams, not only the parts: what passes between components, in what format, under whose control, and what happens when one component fails. Record which component is authoritative for a given decision.
Evidence an auditor or regulator asks for
| A seam map naming the interface, the data passed and the owning team. | Makes the combination inspectable. |
|---|---|
| A fallback test showing behaviour when one component is unavailable. | Gives evidence for the system-level claim. |
| A record of which component is authoritative for each decision. | Answers the accountability question. |
When the model and the rule disagree, which one decides, and who owns that decision?
Common gaps this page catches
- Each component is described separately and the combination is described nowhere.
- No fallback is tested, so the system-level capability is assumed.
- Authority between components is decided in code and recorded in no document.
What to prepare before publication
Draw the seam map for the most complex system in scope. Mark the authoritative component for each decision and test one fallback path.
How to test it here
Take one live system rather than a template, find the field or sentence where Hybrid AI is used, and put this page beside it. Record the answer, the artefact and the date. A glossary entry on its own is an intention; a dated record beside a live artefact is evidence.
What this page does not claim
The page does not reproduce the draft, does not know the descriptor wording that the public preview withholds, and does not certify the system or the organisation. It states the unit at the claim level "methods family" and stops where the public record stops.
Common questions
- What does Hybrid AI mean in ISO/IEC DIS 42102?
- The family covers systems that combine methods: learned components beside rules, symbolic reasoning over a model output, or several models coordinated in one pipeline. The draft also defines conversational learning as a hybrid method that improves through dialogue.
- What evidence should we keep for 5.5?
- A seam map naming the interface, the data passed and the owning team. A fallback test showing behaviour when one component is unavailable.
- Is Hybrid AI a requirement?
- No. The draft is a descriptive framework. It sets no management system requirement and no conformity test. It becomes binding only when a contract, a regulator or a management system adopts it.
Related units
Last updated: · Markdown version · Check your system