There’s something quietly ironic about the federal government using artificial intelligence to sift through public opinion on artificial intelligence.
Artificial Intelligence Minister Evan Solomon has confirmed that an internally developed AI platform is now translating and summarizing more than 11,000 public submissions collected during recent consultations on AI policy. The comments, he says, will be released publicly once the process is complete, and the entire exercise complies with Treasury Board rules.
On the surface, this sounds efficient even sensible. Eleven thousand submissions would take months for human analysts to process, and governments are already notorious for moving at a glacial pace. Using AI to speed things up feels like a practical application of the very technology policymakers are trying to understand and regulate.
But practicality isn’t the only issue here.
Public consultations are supposed to be about listening not just counting keywords or condensing sentiment into neat summaries. When AI becomes the first reader of public concerns, there’s a real question about what gets lost in translation. Nuance, emotional weight, and minority viewpoints don’t always survive algorithmic compression, even when systems are designed with care.
Solomon has emphasized that the system is compliant with Treasury Board rules, and that matters. It suggests safeguards, oversight, and accountability are in place. Still, compliance doesn’t automatically equal trust. For many Canadians, especially those already wary of AI’s growing reach, the idea that software is filtering their voices before policymakers ever see them may feel unsettling.
The government ran these consultations alongside an expert task force tasked with helping update Canada’s national AI strategy. That dual-track approach public input plus expert guidance is smart in theory. The risk is that expert opinion, processed cleanly and efficiently, will outweigh public feedback that arrives summarized, abstracted, and stripped of its raw form.
There’s also a symbolic dimension. If AI is already shaping how public opinion is interpreted, it quietly becomes part of the policy-making process itself. That raises a deeper question: are we consulting Canadians about AI, or are we training AI to consult Canadians for us?
Transparency will be key. Releasing the full comments, not just summaries, is essential if this process is to maintain credibility. Canadians deserve to know that their voices were not merely processed, but genuinely heard.
Using AI to manage scale is understandable. Using it without care risks turning public consultation into a technical exercise rather than a democratic one. As the government shapes the future of AI in Canada, it would do well to remember that efficiency should never come at the cost of trust.
