CAN/DGSI 128, Machine Learning and AI Implementation in Research Institutions

This document specifies minimum requirements for the implementation of machine learning (ML) and artificial intelligence (AI) systems within Canadian research institutions and other institutions that conduct or support research.

This document does not refer to the development of frontier models or advanced AI. However, the practices and guidance outlined here may be useful in supporting a more in-depth standard or policy for AI governance in frontier model or private AI development contexts.

NOTE: Libraries, archives, research centres, labs, universities, colleges and/or CEGEPs and other institutions is not an exhaustive list and any institution, large or small, that determines its work falls under the scope and vulnerabilities described in this standard is encouraged to adopt it.

There are two sub-sections to this standard:
1. Operational Use: Use of AI in the internal systems of research institutions that collect, handle, store, preserve, and disseminate data in support of various research tasks.
2. Research Use: Use of AI tools in the analysis and/or collection and/or preparation of data for research purposes.

NOTE: In both sub-sections the scope covers how AI systems use data, the kinds of functions they can be used for, and basic governance requirements for how AI systems are selected, developed, managed, reviewed, and evaluated.

DATE POSTED: March 20, 2025

DEADLINE FOR COMMENTS: May 19, 2025