Imagine the following scenario. Jordan, a delivery driver in his mid-thirties, is pulled over after a nearby police department’s surveillance system flags his license plate. He’s baffled. The vehicle isn’t stolen, and he hasn’t broken any traffic laws. Minutes later, the officers explain: an AI-powered video analytics surveillance tool had been scanning security footage from a recent break-in, and his route matched one seen in the footage—same neighborhood, similar time of day, comparable van. The match wasn’t 100% exact, but it was enough to trigger an alert.

The officers send him on his way. There’s no ticket, no formal record. Still, the encounter leaves him disconcerted. Every day, he drives the same route. Nothing in his routine has changed, yet he can’t help wondering who’s watching, and what else the system might infer.

Scenarios like this are no longer rare. Across many countries, including Canada and the U.S., police departments are adopting AI tools to process enormous volumes of video from public and private cameras.¹ These systems can scan faces, identify license plates, trace vehicle movements from one location to another, and forecast where a person or vehicle is likely headed next. The idea is to help officers focus their resources and respond to incidents sooner.

AI-driven surveillance now shapes decisions well beyond policing. Schools use it to monitor students’ messages to one another on school-managed platforms.² Hospitals use it to track patient health.³ Corporations use it to watch employee movements.⁴ These tools are appealing to decision-makers and stakeholders because they’re fast, scalable, and reduce human labor. At the same time, their widespread use across industries invites very challenging questions. AI systems can draw from cameras, sensors, social media interactions, GPS records, and biometric data, layering those streams into profiles that follow people across settings. When this type of monitoring hums in the background of our daily lives, how much say do we have over what’s collected, analyzed, or inferred? Do we understand the full scope of what we’ve agreed to, or whether our consent to be monitored was ever truly given? 

Artificial Intelligence (AI): Computational technologies inspired by how humans and other living animals sense, learn, reason, and act, though they often operate in very different ways. AI applications now influence nearly every part of modern society, from defense and national security to civil and criminal justice systems, manufacturing, healthcare, and finance. 

Predictive Policing: The process of taking data from multiple sources, analyzing it, and using the results to anticipate, prevent, and respond to future crime. While intended to improve public safety, predictive policing may also reinforce existing racial and socioeconomic biases in the criminal justice system.

Surveillance Capitalism: A term coined by Harvard scholar Shoshana Zuboff that denotes the widespread collection and commodification of personal data by corporations.⁶ Companies profit by tracking users’ behaviors, predicting future actions, and selling those predictions to advertisers or other third parties.

Biometric Data: Specific physical characteristics such as fingerprints, iris patterns, and facial features that can identify individuals for security or authentication purposes. These identifiers are difficult to replicate, making them valuable for security but risky if exposed.

Algorithmic Bias: Occurs when systematic errors in machine learning algorithms lead to unfair or discriminatory outcomes.⁷ These biases often reflect or reinforce existing socioeconomic, racial, or gender biases present in the data used to train these systems.

The idea of watching as a form of control is hardly new.⁸ In early civilizations, monarchs employed spies, scribes documented enemy movements, and census-takers tracked households—not only to keep records, but to manage taxation, conscription, and social order. Surveillance, even then, rarely operated as a neutral tool. It tended to serve power, often without much resistance.

By the late 18th century, the urge to monitor took physical form. British philosopher Jeremy Bentham sketched a prison unlike anything of its time: the Panopticon, a circular building with a tower at its heart, giving guards the power to watch every cell without being seen themselves.⁹ Centuries later, French theorist Michel Foucault returned to Bentham’s model in his 1975 book Discipline and Punish: The Birth of the Prison.¹⁰ He argued that the Panopticon had become more than a prison—it had become a blueprint for how power could be organized. Bureaucracies like schools, hospitals, and prisons developed their own ways of monitoring behavior, collecting data, and enforcing norms. Surveillance, he claimed, had become a condition of modern life.

These thinkers weren’t writing about AI, yet their frameworks are as relevant as ever.  AI didn’t invent surveillance. It intensified it. What once required a human’s eyes and a paper trail now runs quietly in the background—scanning, categorizing, escalating—before most people even notice that it’s there.

By the early 2000s, digital surveillance had already moved into everyday systems.¹¹ Browsing histories were logged by internet providers. Phones tracked movement by default. Social media harvested behavioral patterns and sold predictions in return. Most people didn’t question it—at least, not out loud. Handing over personal data became routine, and few realized how much information was being collected or why.

That comfort faded as the scale of surveillance came into view.

In 2013, Edward Snowden leaked a series of classified documents exposing the full scope of the U.S. government’s surveillance infrastructure.¹² The National Security Agency wasn’t just targeting suspects. It was collecting email metadata, phone logs, and movement records from millions of people without their knowledge. These systems relied on automated tools that scanned for patterns and anomalies. It wasn’t called AI yet, but the core principles were there: enormous data inputs, algorithmic analysis, and inference without explanation. The revelations sparked international outrage. They also landed at a time when AI was starting to step out of research labs and into everyday life. By the mid-2010s, machine learning and predictive analytics were running recommendation engines, powering facial recognition, and even handling language translation—tools that were soon folded into surveillance in ways that few people noticed at first.¹³

With the rapid rise of AI, new voices started drawing attention to a different kind of watching—one that didn’t involve intelligence agencies at all.

Shoshana Zuboff, a scholar at Harvard Business School, coined the term surveillance capitalism to describe how tech companies profit not from what we do, but from what they can predict we’ll do next.⁶ AI tools, she argued, collect “behavioral surplus”—the leftover traces of our choices, moods, and habits—and turn them into a product. The more the system knows, the more valuable the prediction. 

Kate Crawford, author of Atlas of AI, took the critique one step further. In her view, artificial intelligence is neither magical nor abstract.¹⁴ It’s built from real things—mined resources, labeled images, low-wage labor—and trained on data that reflects deeply uneven histories. Facial recognition doesn’t emerge from nowhere. It learns from photographs, many of which are drawn disproportionately from certain communities. That data becomes the blueprint. Which means, in practice, AI surveillance doesn’t just reflect bias. It reinforces it.

For years, these concerns stayed largely in academic and policy circles. That changed in 2020.

After the murder of George Floyd, protests erupted across the globe. Alongside these calls for racial justice came something less visible, though no less significant: the use of AI surveillance on protestors themselves.¹⁵ Law enforcement agencies had begun using facial recognition tools to identify individuals attending demonstrations. Most of these systems had been adopted with little transparency, and many were trained on datasets known to misidentify people of color.¹⁶ Public trust eroded. Fast. In response, tech giants began to pull back. IBM bowed out of the facial recognition business altogether.¹⁷ Microsoft and Amazon hit pause on selling these tools to law enforcement. 

Policymakers began to stir. 

In 2021, the European Commission introduced the AI Act—the first sweeping attempt to regulate high-risk AI systems.¹⁸ Among the systems classified as “high-risk” were real-time facial recognition tools in public spaces, biometric identification databases, and algorithms used in policing—technologies identified by the European Commission as having the potential to enable mass surveillance, misidentify individuals, and infringe on fundamental rights.

Across the Atlantic, momentum was slower. In the United States, the Algorithmic Accountability Act, first proposed in 2019 and updated in 2023, aimed to require audits of AI systems used in sensitive domains—hiring, housing, credit scoring, and more.¹⁹ The bill promised checks for bias, harm, and discrimination. Two years later, it’s still sitting in committee. 

Across these moments—Bentham’s Panopticon, Snowden’s leaks, the protests of 2020, the European push for reform—a pattern becomes impossible to ignore.  Surveillance has always tracked power. AI didn’t disrupt that tradition. It made it more efficient. What’s changed is how pervasive it’s now become.

Edward Snowden

Former National Security Agency (NSA) contractor and whistleblower Snowden’s 2013 leaks revealed the scope of global surveillance programs operated by the United States and its allies. His disclosures ignited worldwide debates on privacy, civil liberties, and the role of government oversight.

Shoshana Zuboff

One of the first academics to analyze the social, economic, and political consequences of digital technologies, Zuboff is a scholar, author, and former Harvard Business School professor. In The Age of Surveillance Capitalism, she introduces the concept of “behavioral surplus” to describe how tech companies harvest and monetize personal data, turning prediction into profit.⁶

Kate Crawford

A leading scholar on the social implications of artificial intelligence, Crawford is a Research Professor at the University of Southern California and the author of Atlas of AI.¹⁴ Her work focuses on examining large-scale data systems within the broader contexts of history, politics, labor, and the environment.

AI-powered surveillance is rapidly reshaping how information is collected, analyzed, and acted upon, with consequences that ripple across social media platforms, healthcare systems, and financial security.

Surveillance in social media

Social media platforms began as places to chat with loved ones across the globe, post weekend photos, build community, and exchange ideas. They’re still that—but they’ve also become sprawling laboratories for algorithmic surveillance. Every click, scroll, and share becomes training data for AI platforms. The same system that serves you a video “you might like” is also logging your preferences, timing your pauses, and mapping the people you interact with.²⁰

For private companies, that intelligence is gold. Recommendation engines are built to nudge you toward the next purchase, the next article, the next thing that keeps you engaged.²¹ The mechanism is simple: learn your patterns, test a prediction, feed the result back into the model. Over time, the ads, headlines, and trending topics start to feel tailored because they are—each one the product of thousands of tiny inferences.
Governments have learned to tap into the same infrastructure. In the United States, agencies such as the Department of Homeland Security’s  Immigration and Customs Enforcement have purchased access to tools like Babel Street and Fivecast ONYX.²² Babel Street runs large-scale text analytics, scanning public posts, blogs, and forums to gauge sentiment—a measure of whether conversations lean supportive, critical, or hostile toward a given subject. Fivecast ONYX pulls from multiple open sources, including social media, news outlets, and geolocation records, then maps the connections between people and topics. Government contracts describe uses that include monitoring organized crime, protecting national security, tracking gang violence, and investigating cybercrime.

Patient monitoring in healthcare settings

Most patients in hospitals are unattended for much of the day. Nurses spend roughly 37% of their shift in patient care, and physicians average about ten visits per stay.^23,24 That leaves long stretches where staff can’t see what’s happening when a patient is left alone. 

Virtual monitoring systems have tried to bridge that gap. By letting staff observe patients remotely through audio–video feeds, they’ve helped protect high-risk cases. However, they still rely on someone watching in real time, which is nearly impossible to sustain in a busy ward.²⁵ A newer approach uses AI-driven computer vision to detect patterns invisible to routine observation.

One example, the LookDeep Health platform, was tested across 11 hospitals, tracking more than 300 patients at high risk of falling over 1,000 days.²⁶ It identified the presence and roles of people in a room, the layout of furniture, and the frequency of “boundary crossings” such as leaving the bed without assistance.

The AI system’s analysis of mobility, including time spent in bed, in a chair, or walking around, revealed patterns tied to recovery milestones and readiness for discharge. It also flagged patient risk factors, including predictable movement patterns that might end in a fall. As a promising, accessible, non-invasive, and cost-effective form of automated surveillance, these systems may offer a way to improve patient safety and lighten clinician workload.

Uncovering financial crimes

Money laundering is the process of moving illegally obtained money through legitimate channels so its origin may never be traced.²⁷ On paper, it might resemble an ordinary sequence of deposits and transfers. In reality, the activity is often hidden among hundreds of small, scattered transactions—the sort that pass unnoticed until someone starts threading them together.

AI has altered how those threads are discovered. Instead of relying on static rules that flag obvious anomalies, machine learning models can comb through millions of data points to detect behavioral signatures: recurring transfers between seemingly unrelated accounts, transaction timing that mirrors known laundering cycles, or hidden ties to high-risk entities.²⁷ Imagine the difference between checking a single security camera feed and reviewing every camera in a city at once—patterns emerge that would never appear visible in isolation.

HSBC provides a striking example.²⁸ Each month, across 40 million accounts, the bank reviews around 900 million transactions for signs of financial crime. Its Dynamic Risk Assessment platform, built with Google Cloud, uses machine learning to link accounts, chart transaction timing, and pull in relevant information from beyond the bank’s own walls. Rather than flagging a single questionable payment, it reconstructs the network of relationships behind it, offering investigators a far clearer picture. HSBC’s system has identified two to four times more potential criminal activity than traditional approaches while cutting false positives by more than half. That means fewer wasted investigations and more attention on the real threats.

Even so, financial surveillance through AI is far from simple. Banks may still face hurdles in adopting these systems, such as reconciling powerful surveillance capabilities with privacy safeguards.²⁷ As the future unfolds, they’ll need to balance scalable, secure options that modernize traditional IT systems with an equally strong commitment to long-term data security and privacy.

AI surveillance is often promoted as a tool for safety and efficiency, yet in practice, it can spark disputes over accuracy, fairness, and the boundaries of acceptable monitoring. Let’s consider how these tensions arise in schools, public spaces, and the criminal justice system.

School security or student scrutiny?

In school settings, AI-driven surveillance is often pitched as a safety tool designed to catch early signs of violence, depression, student burnout, or self-harm.²⁹ However, its reach can spill far beyond the classroom, scanning private conversations and flagging words without regard for tone, context, or intent.

In 2023, one Tennessee eighth grader learned this the hard way. A joke made in poor taste during an online chat that was typed on a school-monitored platform triggered an automated alert.³⁰ Within hours, she was arrested, strip-searched, and spent the night in a jail cell. The comment was offensive but not a threat. According to the student’s mother, the real harm came from “stupid, stupid technology” that plucked a single word from a conversation and entirely ignored the context.

Software like Gaggle and Lightspeed Alert, which are AI-powered monitoring tools used by school districts, scan student emails, documents, and chat messages in thousands of U.S. schools.³¹ Proponents argue it has saved lives by flagging genuine risks early.² Critics counter that it can criminalize children for careless remarks and deepen mistrust. Research from the Center for Democracy & Technology suggests the impact is uneven—61% of students with learning disabilities say they avoid sharing their true thoughts online because of such monitoring.² This case and these findings indicate that when AI is deputized to patrol student behavior, the line between safeguarding and overreach can be thin. 

Public safety or public targeting?

In the race to design “smart cities,” AI-powered facial recognition has become one of the most talked-about surveillance tools. It watches the flow of people through train stations, stadium gates, and even protest marches, promising that a single glance at the camera can confirm who you are.³² However, its track record tells a different story. Research has found that these tools falsely identify African American and Asian faces ten to one hundred times more often than white faces, with the highest error rates for Native Americans when law enforcement databases were involved.³³ That’s not an abstract statistic; it’s visible on the ground. In Detroit, USA, police records show that in the first half of 2020, face recognition technology was used almost exclusively against Black residents—and in 96% of those cases, it led to misidentification.³⁴ Inaccurate surveillance can lead to devastating consequences, from false arrests to deeper mistrust in the police force.³⁵

Across the Atlantic, the United Kingdom has woven AI-powered facial recognition into many spheres of daily life.³⁶ Cameras scan faces at stadiums, street festivals, protests, and along busy shopping streets, all in the name of public safety—spotting threats, deterring crime, and identifying suspects. Supporters see it as another layer of protection. Critics see something else entirely: when an imperfect system singles out already marginalized groups, it can mean more stops, more profiling, and more harassment, all packaged as “security.”¹⁴ The question then becomes, how much safety is worth that cost?

Is predictive policing reducing crime or trust?

Predictive policing is pitched as a way to get ahead of crime, using algorithms to forecast where incidents might occur or who might be involved.¹ It sounds promising to some: feed the system historical crime data, let it crunch the numbers, and send officers where they’re most “needed.” In practice, it’s far from neutral. Historical data reflects decades of over-policing in Black and other marginalized communities, which means the software may simply automate the same biases under the banner of innovation.¹

Chicago’s Strategic Subject List showed how this can play out. Funded through a federal grant beginning in 2012 and discontinued in 2020, the program used AI to rank people on their “risk” of being involved in gun violence.³⁷ The model looked at past arrests and whether someone was connected—through family, friendship, or a shared police file—to a known shooter or victim. Once the list was built, police and social workers knocked on doors for what they called “pre-crime” interventions. On paper, it was meant to save lives. On the ground, it didn’t change much.

The only measurable pattern was that people flagged by the algorithm were more likely to be arrested for a shooting—a link probably driven by intensified surveillance rather than actual danger.¹ Homicide rates didn’t drop. Critics argue that when the dataset itself is skewed, predictive policing doesn’t correct the problem—it scales it, locking in patterns of over-surveillance for the very communities that have been under the heaviest watch for decades.¹

Inside Amazon’s AI surveillance systems

Few companies embody AI-powered surveillance at work like Amazon. In 2020, the company rolled out an AI tracking system for delivery drivers, combining vehicle sensors, GPS data, and photo verification.⁴ Officially, it was to boost efficiency. Unofficially, it meant every movement could be scrutinized. The system could detect speeding, skipping a seatbelt, or simply pausing too long without touching a package. It wasn’t just keeping tabs—it was judging performance. Miss your productivity target often enough, and the algorithm might generate a warning or even fire you, no manager required.

Warehouses followed a similar playbook. AI tools logged every scan, tracked idle minutes, and calculated whether you were keeping pace with an invisible benchmark. On paper, it looked like precision management. In practice, it meant some employees avoided bathroom breaks entirely for fear of falling behind. At Amazon’s Baltimore facility alone, about 300 workers—over 10% of the staff—were terminated in a single year for “inefficiency.” Multiply that across 75 logistics centers, and the scope becomes clear.

The rising demand for meeting AI-set goals had also taken a toll on employees’ mental health. 55% of workers reported experiencing depression while employed at Amazon.³⁸ More than 80% said they would not apply for a job there again. For the system, these are metrics; for the people behind them, they are human limits stretched thin. AI surveillance in the workplace may promise productivity, but its future will hinge on whether companies can protect efficiency without eroding the mental health of the very people keeping the machine running.

Olympic games mark AI surveillance first in Europe

Months before the Olympic flame reached Paris, another kind of preparation was underway. Law No. 2023-380 authorized what France called “augmented video surveillance”—a trial run for AI systems designed to scan public spaces and flag potential threats in real time.³⁹ The algorithms could track crowd movement, identify unusual patterns, and alert security teams.

Officials stressed that the system avoided biometric tools like facial recognition. Yet the footage could be stored, re-analyzed, and combined with other data sources, making it hard to ignore the reach. 

Groups like La Quadrature du Net, a French digital rights advocacy organization, warned that algorithmic errors might compound the biases already embedded in policing.⁴⁰ The concern was less about a single faulty alert and more about what happens when flawed systems become standard practice.

For organizers, the technology promised smoother security for a high-stakes event. For policymakers, it was a test run for AI surveillance on a national scale. What remains to be seen is the future use of that data—if it will truly be kept for the Olympics only, and never touch the hands of private organizations.

The Dangers of an Artificially Intelligent Future

There’s no doubt we’ve entered the latest revolutionary period in human history: the Technological Revolution. In this article, Eva McCarthy highlights how AI and machine learning algorithms might deepen existing social inequalities. However, when biases are identified and addressed, they hold the potential to foster greater equity in the workplace. 

Ethical AI

With AI now woven into so many aspects of life, it’s essential to prioritize the well-being of both individuals and society. This piece explores what ethical AI means in practice, how it’s shaping industries, and the ongoing challenges of reaching consensus on what counts as ethical design.

1. Fritsch, R. (2025, April). Law enforcement use of AI (AI in Criminal Justice Project, Paper 2). Law Commission of Ontario. https://www.lco-cdo.org/wp-content/uploads/2025/04/LCO-AI-in-Criminal-Justice-Paper-2-Law-Enforcement-Use.pdf
2. Laird, E., Dwyer, M., & Grant-Chapman, H. (2023, September 20). Off task: EdTech threats to student privacy and equity in the age of AI. Center for Democracy & Technology. https://cdt.org/insights/report-off-task-edtech-threats-to-student-privacy-and-equity-in-the-age-of-ai/
3. Solaiman, B., Malik, A., & Ghuloum, S. (2023). Monitoring mental health: Legal and ethical considerations of using artificial intelligence in psychiatric wards. American Journal of Law & Medicine, 49(2–3), 250–266.
4. Bao, Y., Li, W., Ye, Y., & Zhang, Q. (2022, March). Ethical disputes of AI surveillance: Case study of Amazon. In 2022 7th International Conference on Financial Innovation and Economic Development (ICFIED 2022) (pp. 1339–1343). Atlantis Press.
5. Harari, Y. N. (2018). 21 lessons for the 21st century. Spiegel & Grau.
6. Zuboff, S. (2019). The age of surveillance capitalism: The fight for a human future at the new frontier of power. PublicAffairs.
7. Kordzadeh, N., & Ghasemaghaei, M. (2022). Algorithmic bias: Review, synthesis, and future research directions. European Journal of Information Systems, 31(3), 388–409.
8. Lyon, D. (2018). The culture of surveillance: Watching as a way of life. John Wiley & Sons.
9. Bentham, J. (1787). The panoptican writings.
10. Foucault, M. (1975). Surveiller et punir: Naissance de la prison. Gallimard.
11. Monmonier, M. (2004). Spying with maps: Surveillance technologies and the future of privacy. University of Chicago Press.
12. Von Solms, S., & Van Heerden, R. (2015, February). The consequences of Edward Snowden NSA related information disclosures. In ICCWS 2015—The Proceedings of the 10th International Conference on Cyber Warfare and Security: ICCWS2015 (p. 358).
13. Singh, A., Thakur, N., & Sharma, A. (2016, March). A review of supervised machine learning algorithms. In 2016 3rd International Conference on Computing for Sustainable Global Development (INDIACom) (pp. 1310–1315). IEEE.
14. Crawford, K. (2021). The atlas of AI: Power, politics, and the planetary costs of artificial intelligence. Yale University Press.
15. Zalnieriute, M. (2025). Facial recognition surveillance and public space: Protecting protest movements. International Review of Law, Computers & Technology, 39(1), 116–135.
16. Cavazos, J. G., Phillips, P. J., Castillo, C. D., & O’Toole, A. J. (2020). Accuracy comparison across face recognition algorithms: Where are we on measuring race bias? IEEE Transactions on Biometrics, Behavior, and Identity Science, 3(1), 101–111.
17. The Associated Press. (2020, June 11). Microsoft joins Amazon, IBM in pausing face scans for police: Companies end sales of facial recognition technology, despite not being major suppliers. CBC News. https://www.cbc.ca/news/science/face-scans-police-microsoft-1.5608375
18. Gesley, J. (2021, May 26). European Union: Commission publishes proposal to regulate artificial intelligence. Library of Congress. https://www.loc.gov/item/global-legal-monitor/2021-05-26/european-union-commission-publishes-proposal-to-regulate-artificial-intelligence/
19. Congress.gov. (2023a, September 21). S.2892 — Algorithmic Accountability Act of 2023. United States Congress. https://www.congress.gov/bill/118th-congress/senate-bill/2892
20. Anandhan, A., Shuib, L., Ismail, M. A., & Mujtaba, G. (2018). Social media recommender systems: Review and open research issues. IEEE Access, 6, 15608–15628.
21. Jesse, M., & Jannach, D. (2021). Digital nudging with recommender systems: Survey and future directions. Computers in Human Behavior Reports, 3, 100052.
22. U.S. Department of Homeland Security. (2024, December 16). AI at DHS: A deep dive into our use case inventory. https://www.dhs.gov/archive/news/2024/12/16/ai-dhs-deep-dive-our-use-case-inventory
23. Westbrook, J. I., Duffield, C., Li, L., & Creswick, N. J. (2011). How much time do nurses have for patients? A longitudinal study quantifying hospital nurses' patterns of task time distribution and interactions with health professionals. BMC Health Services Research, 11, 1–12. https://doi.org/10.1186/1472-6963-11-319
24. Chae, W., Choi, D.-W., Park, E.-C., & Jang, S.-I. (2021). Improved inpatient care through greater patient-doctor contact under the hospitalist management approach: A real-time assessment. International Journal of Environmental Research and Public Health, 18, 5718. https://doi.org/10.3390/ijerph18115718
25. Abbe, J. R., & O'Keeffe, C. (2021). Continuous video monitoring: Implementation strategies for safe patient care and identified best practices. Journal of Nursing Care Quality, 36(2), 137–142.
26. Gabriel, P., Rehani, P., Troy, T., Wyatt, T., Choma, M., & Singh, N. (2025). Continuous patient monitoring with AI: Real-time analysis of video in hospital care settings. Frontiers in Imaging, 4, 1547166.
27. Han, J., Huang, Y., Liu, S., & Towey, K. (2020). Artificial intelligence for anti-money laundering: A review and extension. Digital Finance, 2(3), 211–239.
28. Calvery, J. (2024, June 10). Harnessing the power of AI to fight financial crime. HSBC. https://www.hsbc.com/news-and-views/views/hsbc-views/harnessing-the-power-of-ai-to-fight-financial-crime
29. Ayer, L., Boudreaux, B., Paige, J. W., Holmes, P., Blagg, T. L., & Mendon-Plasek, S. J. (2024). Artificial intelligence–based student activity monitoring for suicide risk: Considerations for K–12 schools, caregivers, government, and technology developers. Rand Health Quarterly, 11(2), 2.
30. Lurye, S. (2025, August 7). Schools are using AI to protect students. It also leads to false alarms — and arrests. ABC News. https://abcnews.go.com/Technology/wireStory/schools-ai-surveillance-protect-students-leads-false-alarms-124440368
31. Rogerson, K., & Sherman, J. (2021). AI in public education: Humble beginnings and revolutionary potential. In Towards an international political economy of artificial intelligence (pp. 63–83). Springer International Publishing.
32. Tao, M., Jiang, R., & Downs, C. (2022). Ethics of face recognition in smart cities toward trustworthy AI. In Big data privacy and security in smart cities (pp. 23–52). Springer International Publishing.
33. Grother, P., Ngan, M., & Hanaoka, K. (2019). Face recognition vendor test (FVRT): Part 3, demographic effects. Gaithersburg, MD: National Institute of Standards and Technology.
34. Koebler, J. (2020, June 29). Detroit police chief: Facial recognition software misidentifies 96% of the time. Vice. https://www.vice.com/en_us/article/dyzykz/detroit-police-chief-facial-recognition-software-misidentifies-96-of-the-time
35. Allyn, B. (2020, June 24). “The computer got it wrong”: How facial recognition led to false arrest of Black man. NPR. https://www.npr.org/2020/06/24/882683463/the-computer-got-it-wrong-how-facial-recognition-led-to-a-false-arrest-inmichig
36. Fielddrive. (2024, November 26). The impact of facial recognition at events in UK: A closer look. https://www.fielddrive.com/blog/facial-recognition-at-events-uk-challenges#facial-recognition-at-events-in-uk-case-studies
37. Stroud, M. (2016, August 19). Chicago’s predictive policing tool just failed a major test: A RAND report shows that the ‘Strategic Subject List’ doesn’t reduce homicides. The Verge. https://www.theverge.com/2016/8/19/12552384/chicago-heat-list-tool-failed-rand-test
38. UNI Global Union. (2023, January). Life in the Amazon Panopticon: An international survey of Amazon workers. Jarrow Insights for UNI Global Union. https://uniglobalunion.org/wp-content/uploads/UNIAMZN_Report.pdf
39. Buchanan, K. (2023, November 15). France: New law establishes legal framework for 2024 Olympic and Paralympic Games. Library of Congress. https://www.loc.gov/item/global-legal-monitor/2023-11-15/france-new-law-establishes-legal-framework-for-2024-olympic-and-paralympic-games/
40. La Quadrature du Net. (2024, May 2). Against the empire of algorithmic video-surveillance, La Quadrature du Net strikes back. https://www.laquadrature.net/en/2024/05/02/against-the-empire-of-algorithmic-video-surveillance-la-quadrature-du-net-strikes-back/