- Formalizing Quality Metrics and Lifecycle Processes for Synthetic Data and ISO/IEC AWI TR 42103
Artificial Intelligence
In the medium term, providers of high-risk AI systems are expected to obtain a common way to evidence AI Act Art. 10, 11/Annex IV and 13 for synthetic and augmented data, generation intent, provenance, generator card, validation report, release label and chain of custody, and auditors a common object of examination; SMEs, who cannot build custom validation frameworks, are expected to get a tiered path in which an exploratory dataset needs a page and a conformity-critical dataset the full report. In the longer term, a WG 2 Technical Specification that JTC 21's prEN 18284 can reference, an EHDS secondary-use instrument with a quality layer, clinical AI trained on augmented data that does not entrench bias against under-represented groups, and carbon cost of generation made visible. For Europe, the 56 European projects form the evidence base of an SC 42 roadmap item, and I am now a Hungarian expert who sits in the national AI committee that will carry the positi
- New standard development
- Revision of existing work items
-
My activity targets ISO/IEC AWI TR 42103 (moving toward CD), “Information technology - Artificial intelligence - Overview of synthetic data in the context of AI systems”, under ISO/IEC JTC 1/SC 42. The contribution targets WG 1 (Foundational standards, responsible for TR 42103, terminology and lifecycle) and WG 3 (Trustworthiness), where a multi-dimensional Quality Assurance and risk framework is intended for submission as an informative annex. National participation is channelled via the Hungarian Standards Body (MSZT), national mirror committee MSZT/MB 819 Informatics.
Title & Organisation Name: 4D Consulting Kft.
Country: Hungary
