AI 'Robot MP' Created by Build Canada Raises Debate Over Technology’s Role in Policy-Making

Shortly after the most recent Canadian federal election, the advocacy group Build Canada introduced an initiative it described as the world's first artificial intelligence member of Parliament. The so-called "Builder MP" project is based on OpenAI’s GPT-5 and is designed to review every piece of legislation introduced in the House of Commons. The system summarizes each bill, analyzes it according to a set of criteria reflecting Build Canada’s economic freedom-oriented policy objectives, and assigns it a vote—either yes, no, or abstain. The AI MP also generates proposed questions in the style of parliamentary question period and occasionally suggests amendments to legislation.

Build Canada CEO Lucy Hargreaves described the AI MP as an illustration of how technological advancements are poised to modernize public life and reshape the functioning of democracies. "This project is about showing that the same technologies transforming industries can also modernize our public life.... AI will reshape how societies function, including how democracies govern themselves," Hargreaves wrote at launch. The framing positions AI as an inevitable force in public administration, echoing broader trends in government adoption of artificial intelligence.

OpenAI, the developer of GPT-5, has been the subject of recent scrutiny in Canada. The federal privacy commissioner determined that OpenAI had violated federal and provincial privacy laws during the training of its ChatGPT system. Additionally, the company is facing litigation from families in Tumbler Ridge, B.C., who allege that OpenAI failed to notify authorities about a planned mass shooting that it had knowledge of. The Canadian government’s recently announced online harms legislation did not include a mandatory reporting requirement for similar situations, a decision criticized by B.C. Premier David Eby.

Globally, governments are experimenting with AI integration. Albania has appointed an AI cabinet minister; Sweden’s prime minister consults AI for policy opinions; the United Kingdom uses AI for local planning decisions; and the European Union employs AI to evaluate the legal compatibility of prospective member countries. In Canada, AI was used to review and summarize public submissions during consultations on a national AI strategy—a task typically performed by non-partisan public servants. Federal AI Minister Evan Solomon disclosed that he had used AI for a legislative briefing, which reportedly resulted in significant errors.

Since the formation of the new Liberal government, Build Canada’s AI MP has reviewed and cast votes on approximately 200 bills. The tool’s performance has faced criticism for factual inaccuracies, inconsistencies, and analytical limitations. For example, in its analysis of Bill C-253, which proposes a national framework for a guaranteed livable basic income for individuals over 17, the AI MP incorrectly stated the bill would deliver basic income to "all people in Canada aged 17 and over," misrepresenting the bill’s scope. Such mischaracterizations could have significant policy and fiscal implications if acted upon.

The AI MP's voting patterns on recognition and designation bills have also shown inconsistency. It abstained on the "Ukrainian Heritage Month Act" as well as acts recognizing Arab, Caribbean, and Somalian heritage months, but voted no on bills for Albanian heritage, Hellenic heritage, and national immigration recognition. The bills share nearly identical language, yet the AI MP’s rationales for divergent votes are substantively similar, raising questions about the tool’s internal logic and reliability.

In some instances, the AI MP’s proposed questions for debate do not align with the stance it adopts on a bill. For example, it recommended a yes vote on a private member's bill for a U.S.-style "castle law" but framed its questions about the bill as if it opposed it. Such discrepancies highlight challenges in ensuring coherence between AI-generated analysis and policy recommendations.

Build Canada’s AI MP is explicitly programmed to align with the organization’s economic freedom agenda. This orientation is evident in its voting and commentary. The tool supported several tough-on-crime bills, including Bill C-14 (bail and sentencing reform) and the aforementioned "castle law," but abstained on legislation proposing tougher sentences for domestic violence, citing concerns about expanding state control rather than promoting economic liberty. On the Gatineau Park Act—intended to protect and preserve public access—the AI MP recommended implementing dynamic pricing to fully fund upkeep, a position with potential benefit to Build Canada's interests. The tool has suggested bill revisions that would directly serve the policy preferences of Build Canada’s membership on multiple occasions.

Regarding Indigenous issues, the AI MP abstained or voted no on several bills addressing First Nations treaties, Indigenous rights, and residential school denialism, arguing that some measures risked a "chilling effect on discourse." It supported the First Nations Clean Water Act, though the Assembly of First Nations criticized the bill for having weaker protections compared to earlier versions.

On major government legislation, the AI MP supported the One Canadian Economy Act (Bill C-5), which repeals carbon pricing and introduces tax cuts, in addition to the government’s two primary budget bills. It opposed the digital harms bill and most climate-related bills, including those focused on renewable energy, climate finance, and ocean pollution. The AI MP also voted against the groceries benefit program and opposed border and immigration bills (C-2 and C-12), not for civil rights reasons but due to concerns about regulatory complexity and perceived costs to exporters, financial institutions, and employers.

Some of the AI MP’s decisions have been viewed as lacking practical or emotional sensitivity. In one case, it voted against a private member’s bill to extend employment insurance and bereavement leave to parents grieving the loss of a child, despite unanimous parliamentary support at second reading. The AI MP justified its opposition by stating that the bill was neither a major prosperity reform nor a low-cost change. It also voted against bills creating a national framework for attention deficit hyperactivity disorder, a national strategy for children and youth, and a bill supporting homeless Canadians, explaining that such measures did not advance growth, productivity, or competitiveness.

Notably, the AI MP voted against all routine treasury bills that provide operational funding for government, a pattern that—if followed by a real government—would result in repeated losses of parliamentary confidence and trigger frequent elections.

The emergence of Build Canada and its AI MP project has prompted debate about the influence of technology-focused advocacy groups on policy-making and the meaning of "building Canada." While the phrase is used by the Mark Carney government in communications and could refer to a range of nation-building efforts, it has also become associated with a movement advocating rapid technological adoption, sometimes criticized for insufficient regard for democratic norms, human rights, social justice, or environmental considerations. The "Techno-Optimist Manifesto," authored in 2023 by U.S. entrepreneur Marc Andreessen, concluded with the call, "It’s time to build," reflecting an ethos of unrestrained technological progress.

As Build Canada and its AI-driven policy analysis gain visibility and legitimacy within Canadian public policy discussions, questions have arisen regarding the transparency and public benefit of their recommendations. The group’s proposals, their beneficiaries, and their necessity are now under heightened scrutiny from political stakeholders, advocacy groups, and the general public. The debate reflects broader tensions regarding the integration of advanced technologies into democratic governance and the need to balance innovation with accountability and representation.