Canadian federal departments are using generative AI without a common binding framework covering its use, disclosure and accountability, The Walrus reports. The gap matters most when AI-assisted work can affect people’s access to public services.
Watch Desk analysis
What happened
In a report published on 10 November, journalist Colin Horgan describes uneven oversight of AI across federal departments. The article says the existing Directive on Automated Decision-Making applies to certain automated decisions but was not designed for generative AI tools that public servants may use for tasks such as research, summarising and analysis.
The Walrus says Immigration, Refugees and Citizenship Canada confirmed that staff use publicly available generative AI tools, but would not identify them for the article, and that training on those tools is not mandatory. The report also cites a KPMG survey of public-sector workers across federal, provincial and municipal levels: nearly half of respondents said they frequently use AI for work. That survey is not a measure of federal employees alone.
Why it matters
When public servants use generative AI around applications, benefits or other consequential work, errors and unclear responsibility can affect people dealing with the state. The article recounts a rejected permanent-residence application in which a refusal letter included job duties the applicant said were not hers. It says the letter disclosed AI had been used in processing, while IRCC maintains that a person makes the final decision. That is a serious account, not proof that AI alone caused the rejection.
The wider concern is that rules for some automated decisions may not cover the less formal, everyday use of generative tools. Without a clear view of which tools are used, for what work and under what safeguards, it is harder for the public to understand how a decision was reached or how to challenge an error.
Our read
The useful question is not whether government should use AI at all. It is whether public bodies can explain where it is used, train staff to handle it, protect sensitive information and make a named person answerable for consequential decisions. The Walrus’s reporting points to gaps on each front. “A human makes the final call” is a start, not a complete account of what happened before that call.
What to watch
- Whether Ottawa sets binding rules specifically for generative AI use in federal departments.
- Whether departments disclose which tools staff use and for which tasks.
- Whether mandatory training and clear routes to challenge AI-assisted decisions follow.
Discussion spark: Should federal departments be required to disclose every use of generative AI in work affecting people’s applications or benefits, or is that too broad to be practical?
Sources and evidence
- As Ottawa Gambles on AI, Canadians Live with the Mistakes (10 November 2026, 11:30 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.