Ottawa Has a New Tool Against Election Lies. A Privacy Lawyer Isn’t Sure the Cure Is Worse Than the Disease.
Bill C-25 targets AI-generated disinformation and foreign funding loopholes in Canadian elections law. But a privacy lawyer told the committee the cure may be worse than the disease.
The problem arrived before any witness said a word. Ask an AI chatbot a simple question about where to vote or whether your ballot is secret, and there is roughly a one-in-three chance it will give you a factually wrong answer. That was the finding of a UK-based think tank cited before the Standing Committee on Procedure and House Affairs during study of Bill C-25, and it framed every argument that followed. Elections, the committee was told, are no longer just vulnerable to foreign cash and partisan spin. The threat now hallucinates.
Bill C-25, described in committee testimony as a “security patch” for Canadian elections, targets two specific vulnerabilities: false statements about candidates and voting that may be amplified or generated by artificial intelligence, and loopholes in third-party financing rules that allow organizations to mix money before an election year and then spend it in ways that obscure the original source. Witnesses and committee members described AI systems that can already generate plausible disinformation at scale and, more troublingly, AI agents capable of operating semi-autonomously in ways that may allow them to navigate around existing campaign finance restrictions.
The case for the bill, as laid out before the committee, was straightforward: existing elections law was written in an era when a human had to decide to tell a lie and find a way to publish it. That friction is gone.
“The Cure Is Worse Than the Disease”
Not every witness agreed the proposed fixes were sound.
Gerald Chipeur, whose testimony focused on proposed privacy amendments to the Canada Elections Act, offered the sharpest objection on the public record.
“The cure is worse than the disease,” Chipeur told the committee. “The publication of voter names and addresses has never been a problem in the past.”
The remark crystallized a tension running through the committee’s study. Several proposed amendments to election law touched the voter registry and the disclosure of elector information. Critics of those provisions argued that making voter data more broadly available, or altering the conditions under which it circulates, introduces a privacy risk that far exceeds whatever disinformation problem it purports to solve. For Chipeur, that was the crux: the existing system, whatever its shortcomings, had not produced documented harm from the publication of voter identifying information. The remedy was being prescribed for a hypothetical.
That argument did not go unchallenged. Proponents of reform pointed out that the environment in which voter data circulates has changed fundamentally. A list of names and addresses that once required physical distribution and manual use can now be cross-referenced, enriched, and deployed at a volume and speed that the original drafters of the Canada Elections Act could not have anticipated.
The 10 Percent Loophole
On the financing side, the committee examined a structural vulnerability in the existing rules governing third-party spending.
Under the current framework, organizations that receive funding from foreign or undisclosed sources are subject to restrictions on how much of that money can be directed toward election-related activity. But the rules are calibrated around activity that happens during a defined pre-election or election period. Witnesses told the committee that organizations can mix revenues well before that window opens, in amounts that remain under the 10 percent threshold that would otherwise trigger disclosure requirements. Once the election period begins, that pre-mixed funding is available and its origins are, for practical purposes, obscured.
The effect, as described in testimony, is that the financing rules create an incentive to front-load influence spending in a way that leaves the public record with no clear picture of where the money came from. Bill C-25 was intended, in part, to address this by tightening the conditions under which pre-election revenue can flow into election-period spending.
Privacy and Political Parties
A related thread in the committee’s study concerned data privacy within political parties themselves.
Unlike most organizations that collect and process personal information, federal political parties are not subject to standard provincial privacy legislation. Witnesses argued before the committee that this exemption was not justified on policy grounds and that parties should be brought within the standard privacy framework that governs comparable institutions. The committee heard that Canadian voters have limited visibility into how their personal information is collected, retained, and used by the parties that seek their support.
The AI dimension reinforced this concern. If AI systems are now capable of processing and acting on voter data in real time, the question of what data parties hold and what they are permitted to do with it takes on a different weight than it did when canvassing meant a volunteer with a clipboard.
What the Record Shows
The committee produced no settled resolution of these debates. The Chipeur objection stood alongside the AI threat evidence without either overriding the other. That is, in a sense, the honest state of the record: a parliament trying to patch rules written for a pre-AI election environment, confronting genuine uncertainty about whether the patches might introduce problems as serious as the ones they fix.
The 34 percent error rate on AI chatbot responses about electoral information was not a hypothetical. It was a documented finding, placed in evidence before the committee. For a country in which nearly every federal election in living memory has been decided by margins of a few percentage points in a few dozen ridings, that number deserves to sit with a reader for a moment.
The “security patch” framing, offered in testimony, was not boastful. It was modest: an acknowledgment that the legislation was not a solution, but a repair. And repair work, as the Chipeur testimony made plain, carries its own risk of breaking something that currently works.
Hansard Files spends weeks in the primary record so you don’t have to. If this kind of accountability coverage matters to you, consider subscribing to keep it independent.
Source Documents
Standing Committee on Procedure and House Affairs. (2026). Evidence (PROCEV34-E). House of Commons of Canada, 45th Parliament, 1st Session.
Standing Committee on Procedure and House Affairs. (2026). Evidence (PROCEV35-E). House of Commons of Canada, 45th Parliament, 1st Session.
Standing Committee on Procedure and House Affairs. (2026). Evidence (PROCEV36-E). House of Commons of Canada, 45th Parliament, 1st Session.