The Myth of Neutral Innovation: Why Canada's AI Strategy Risks Codifying Gender Bias

The prevailing narrative surrounding artificial intelligence is one of neutral progress—a tool that, if deployed efficiently, elevates productivity for all. However, as Canada moves forward with its national AI strategy, we must confront a systemic failure: the delusion that technology is a blank slate. In reality, AI is a mirror. If the data we feed it is skewed, the resulting systems will not only reflect those biases but will institutionalize them.

As CBC News first reported, the Canadian government unveiled its "AI for All" national strategy in June, backed by a $2-billion plan intended to update laws and protect vulnerable groups from algorithmic biases and online violence. While the commitment of funds is a start, the framework remains dangerously superficial if it does not address the root of the problem: who is building these systems and what data are they using?

Indra Kubicek, CEO of the Montreal-based charity Digital Moment, hit the mark during a House of Commons status of women committee meeting when she stated that AI is not neutral. Because these systems learn from historical data, they inherently reflect existing inequalities. Kubicek warned that without proper safeguards, we risk codifying these biases into the very systems that influence our daily lives. The danger is not a lack of ability among women or 2SLGBTQ+ individuals, but a failure to create the conditions for their success in AI development and governance.

This is not a theoretical concern; it is an economic and physical risk. A March report from the United Nations' International Labour Organization (ILO) indicated that generative AI has a disproportionate impact on industries dominated by women. The ILO found that approximately 29 percent of female-dominated jobs are exposed to generative AI, compared to just 16 percent of male-dominated roles. Women are also over-represented in high-risk automation categories, such as administrative and clerical roles. When the government speaks of "softening transition shocks" in these sectors, it must acknowledge that the shock is gendered.

Furthermore, the risk extends into the realm of survival. Amy Flood, executive director of Women's Health Collective Canada, testified to the parliamentary committee that AI risks making "already invisible populations even more invisible." Flood noted that for decades, women have been missing from significant portions of health research. If AI is built on this incomplete data, it will reinforce existing inequities, leading to poorer health outcomes for women and gender-diverse populations. As Flood rightly argued, we cannot build equitable systems using data that fails to represent the people it is meant to serve.

Even the UN Committee on the Elimination of Discrimination against Women (CEDAW) has issued a warning that AI systems may amplify gender stereotypes and discrimination, particularly in critical areas like health care, education, and employment. While Prime Minister Mark Carney has joined other world leaders in acknowledging the need for "strengthened oversight" due to safety and security risks, the focus remains too broad.

To avoid a future where innovation is merely a tool for marginalization, Canada needs more than a $2-billion budget; it needs a national data strategy that prioritizes representation. If the architects of AI continue to ignore the gender-coded biases of the past, the "AI for All" strategy will be a misnomer. It will not be AI for all—it will be AI for those already in power, automated at scale.