When Safety Becomes a Surveillance Slippery Slope: Montreal’s AI Experiment
Imagine walking down a street where every move is logged—not by a human officer, but by a network of AI-powered cameras silently scanning your face, your car, even the color of your jacket. This isn’t a dystopian novel; it’s Montreal’s evolving reality. The city’s police force, the SPVM, has quietly expanded its use of AI surveillance tools, sparking urgent debates about privacy, oversight, and the very definition of public safety. But here’s the uncomfortable truth: we’re sleepwalking into a surveillance state, and the line between protection and intrusion has never been blurrier.
The Illusion of Transparency
Let’s start with the most glaring hypocrisy. Montreal’s police chief initially claimed ignorance about a Cloudrunner camera spotted on a lamppost—a device designed to scan vehicles and store data on their make, model, and color. When confronted, the SPVM backpedaled, calling it a temporary measure for “targeted investigations.” But if the chief didn’t know about a tool his own department deployed, what else are they hiding? This isn’t just incompetence; it’s systemic opacity. Police argue they’re protecting investigative strategies, but in reality, they’re exploiting a loophole: secrecy shields accountability. What many people don’t realize is that this lack of transparency isn’t unique to Montreal. From London’s facial recognition trials to Chicago’s predictive policing algorithms, authorities worldwide weaponize complexity to avoid scrutiny. The result? A public kept in the dark about how their data is harvested and used.
The ‘Red Coat’ Defense: A Flawed Safeguard
The SPVM insists its AI software isn’t used for facial recognition. Instead, it tracks objects like “red coats” or “green bikes” to avoid targeting individuals based on ethnicity or other biometric traits. In theory, this sounds like a reasonable compromise. But let’s unpack this. First, object-based tracking still creates a de facto surveillance net. A “red coat” might be worn by someone attending a protest, visiting a specific location, or simply walking home. Second, this logic assumes AI systems are neutral arbiters of “objects,” ignoring how bias seeps into algorithms. For instance, a camera trained to flag “suspicious” vehicle patterns might disproportionately target low-income neighborhoods where older cars are more common. The SPVM’s 2023 privacy assessment admits this risk but downplays it, claiming oversight protocols will prevent discrimination. Personally, I find this naïve. Technology doesn’t exist in a vacuum—it reflects the priorities of its creators. Without independent audits, these “safeguards” are just wishful thinking.
The Profit Motive Behind Public Safety
Here’s a detail that deserves more scrutiny: Montreal awarded Genetec, the Cloudrunner manufacturer, a $17,134 contract for four units. Genetec’s vice-president argues the tech “respects privacy” because data interactions are logged. But let’s follow the money. Companies like Genetec, Rank One Computing, and Clearview AI are in the business of selling surveillance solutions. Their incentives? Sell more tools, expand their reach, and lobby against regulations. This isn’t hypothetical—Clearview AI faced lawsuits in Canada and the U.S. for scraping billions of facial images from social media. Yet, cities keep partnering with them, often under the guise of “fighting crime.” What this really suggests is a symbiotic relationship between law enforcement and tech firms, where public safety becomes a marketing slogan for profit-driven tools. And when these systems fail—or worse, enable abuse—taxpayers foot the bill.
The False Choice: Safety vs. Privacy
The SPVM frames its AI rollout as a necessary trade-off: surrender some privacy to combat vehicle thefts, armed violence, and merchant crimes. But this binary framing is dangerously simplistic. First, does mass surveillance even work? Studies in cities like Detroit and London show facial recognition systems have high error rates, often misidentifying people of color. Second, what are the hidden costs? When every citizen is a potential suspect, trust in institutions erodes. I’ve spoken to Montreal residents who now avoid certain neighborhoods, fearing they’ll be profiled by a camera. Third, the “safety” argument ignores root causes: poverty, mental health crises, and systemic racism. Pouring resources into AI surveillance instead of social programs is like using a tourniquet on a bullet wound. The deeper issue? We’re outsourcing complex societal challenges to technology that cannot—and should not—replace human judgment.
A Call for Democratic Oversight
The Ligue des droits et libertés, a Quebec civil liberties group, rightly demands transparency. But asking police to self-regulate is like letting a fox guard the henhouse. What’s needed is independent oversight—elected officials, civil society groups, and technologists working together to define limits. Why not a public registry of all surveillance tools? Or a moratorium on AI deployments until ethical guidelines are established? The Commission d’accès à l’information’s requirement that public surveillance must address a “legitimate, urgent objective” is a start, but enforcement is toothless without consequences. From my perspective, the solution lies in democratizing surveillance decisions. If citizens can’t vote on how their data is used, they’ll never reclaim their right to exist without being watched.
The Crossroads of Technology and Trust
Montreal’s AI surveillance experiment isn’t just about cameras on lampposts—it’s a microcosm of a global reckoning. Will we accept a future where algorithms dictate our freedoms, or will we draw a line in the digital sand? The answer hinges on one question: Who serves whom? If technology exists to empower authorities without accountability, we risk becoming prisoners of a system we never agreed to. But if we demand transparency, equity, and human dignity as non-negotiables, there’s still time to steer toward a safer, freer world. The choice, however, is slipping away faster than we think.
