Hide in Plain Sight: Clothing as Resistance to AI Surveillance

By Javier Surasky

This is the original post version
A Spanish (ES) version is available here

People observed by surveillance cameras symbolizing anti-surveillance fashion and AI-driven algorithmic control


Anti-surveillance fashion

Visibility is no longer experienced only as a social experience. It is increasinglyadministered as a technical regime. Cameras, sensors, and artificial intelligence systems monitor bodies continuously through automated processes that run in the background and, in many cases, remain unnoticed.

Within this landscape, anti-surveillance fashion has emerged as a still marginal practice. It relies on clothing, makeup, and accessories designed to disrupt algorithmic surveillance and facial recognition, especially systems grounded in computer vision and machine learning.

From this standpoint, anti-surveillance fashion should be read less as an aesthetic trend than as a reaction to the expansion of AI as an infrastructure of control. For that reason, it is not best understood as a technical fix to surveillance, but rather as a political and social indicator. Its appearance points to a deeper structural shift: surveillance logics are moving away from institutionalarchitectures and reattaching themselves to the individual body as a new anchor point.

Algorithmic surveillance

Contemporary surveillance does not depend so much on human operators as on AI, which observes without fatigue, classifies without context, and decides without ethics: “AI-driven surveillance systems have been widely deployed across human-centered public spaces, including schools, commercial facilities, and urban roadways” (Long et al., 2025:1), creating an ecosystemic scenario of automated surveillance.

In Foucaultian terms, one way to situate this turn is to start from practices rather than from the “universals” (State, society, sovereignty) that usually organize our explanations: if we assume “that universals do not exist” (Foucault, 2007:18), algorithmic surveillance becomes intelligible as a set of procedures that produce social reality.

Instead of understanding people as subjects, systems tend to reduce them to patterns: “The datafication of the body separates the physical from the digital, turning embodied subjects into information that can be transmitted electronically” (Rogers, 2025:24), which becomes a source of data from which identities, intentions, or dangerousness are extracted. Algorithmic surveillance does not need to know who you are, but which category you fall into, or, in other terms, it does not recognize subjects but correlations.

Clothing, which historically has operated as a cultural and identity marker, also becomes a variable of defense: texture, color, form, and movement are integrated into that flow of data that feeds AI models, in order to confront them on their own terrain.

Lessons from the First World War

Adam Harvey’s work on CV Dazzle is an unavoidable antecedent within the topic we analyze: it is an application of First World War “dazzle” camouflage to computer vision. Dazzle camouflage did not seek to make objects such as warships invisible, but rather to confuse the enemy’s calculations about course, speed, and orientation by applying high-contrast colors and geometric paint patterns.

Harvey applies that logic of perceptual disruption to the human face through high-contrast, asymmetrical makeup and hairstyles to break the configuration that algorithms expect to detect in a face, since, if there is no detectable face, the rest of the biometric analysis is not activated. These aesthetic interventions make it possible to deceive systems through an “open-source anti-facial recognition toolkit where hair and make-up can be styled to camouflage facial features from facial detection software programs” (Calvi, 2023:83).

Makeup is no longer oriented toward “looking good,” but toward generating a “form of unrecognisability” (Calvi, 2023:83), an invisibility in full exposure, which connects with a broader genealogy of anti-surveillance objects and practices, where the aesthetic functions as an interface of resistance.

The body as a surface for data capture

With the term “biopolitics,” Foucault names a specific form of governmental problematization of life: “I understood it as the way in which, since the eighteenth century, attempts have been made to rationalize the problems posed to governmental practice by phenomena characteristic of a set of living beings constituted as a population: health, hygiene, birthrate, longevity, races” (Foucault, 2007:359).

However, as Lemke warns (2017, 14–15), the concept of biopolitics has a range of possible interpretations and, consequently, cannot be applied automatically; it requires an operation of conceptual precision.

Given that AI redefines the status of the body, now become a biological interface from which vectors are generated, an opposition is produced between a situated and vulnerable human body and an omnipresent and powerful system that observes it. This mismatch gives rise to anti-surveillance fashion as a strategy for repoliticizing the body and its role as an agent in the face of an environment that appropriates it as a passive input.

Dressing against surveillance is an attempt to break the fluidity of a system by introducing visual noise where order is presumed: anti-surveillance camouflage allows people to hide in plain sight, but it has an insurmountable limit insofar as it proposes an individual exit (Monahan, 2015), which affects its political weight. But to understand it fully, we must understand how anti-surveillance fashion operates.

Four forms of resistance through clothing

Anti-surveillance fashion is based on four broad groups of intervention that share the same premise: the adversary is no longer a human observer, but a model trained on massive data.

a) Algorithmic interference: The use of visual patterns designed to confuse computer vision models, since “even small pixel-level changes or subtle texture manipulations can trigger severe detection failures” (Zhou et al., 2025:1).

b) Sensory blocking: Materials that affect infrared sensors, thermal cameras, or depth systems. In the case of thermal surveillance, for example, “it is required that this piece of clothing will have a certain adversarial effect from any angle” (Zhu et al., 2021:3), in such a way that it disrupts the system’s perception of angles, distances, and/or movements.

c) Critical urban camouflage: Kronman (2023:17) formulates it in playful terms: “playing the game of avoidance and tricking AI with anti-surveillance designs is a type of urban hide-and-seek.” Urban camouflage seeks a tactical reconfiguration of presence: moving, blending in, and diverting algorithmic reading in an environment where the background itself is already detection infrastructure.

d) Symbolic disruption: These are designs that are not oriented toward evading surveillance, but toward making it visible. Clothing functions as public denunciation, not as a shield.

Technical limitations of anti-surveillance fashion

From a strictly technical point of view, the response works poorly and inconsistently. Systems evolve, adapt, retrain, and operate in multimodal ways, so surveillance does not depend on a single signal.

But the technical field of countersurveillance is advancing: at the beginning, external and specific elements were used, from lasers to light-emission devices that had to be aimed at cameras. Today, the introduction of visual noise in clothing is privileged, making it unnecessary to carry any other equipment. So-called adversarial attacks, for example, operate by segmenting people through transformations and the occlusion of “areas of clothing in images” (Treu et al., 2021:3). Each garment becomes a support for deforming the image and, at the same time, its use can be justified as an ordinary aesthetic choice.

There are even garments with controlled activation: “By leveraging temperature as a control signal, the system simultaneously activates both RGB and infrared patches, thereby achieving dual-modal evasion in a controllable manner” (Long et al., 2025:8).

None of this changes the reality that anti-surveillance techniques do not compete on equal terms with state or corporate infrastructures and, worse still, anti-surveillance tactics may even amplify factors of discrimination: “this aestheticization of resistance and its accompanying discourses have serious blind spots, specifically where issues of racial identity, difference, and power are concerned. Given that biometric systems already ‘fail’ at a greater rate for racial minorities, effectively nominating those populations for increased scrutiny, what might be the effects of someone marked as Other openly and intentionally challenging state surveillance systems?” (Monahan, 2015:165).

The garment that protects also confirms the rule: the surveillance state will remain omnipresent, and it creates a paradox, since the very practice that makes the problem visible contributes to fragmenting the possibility of a collective response.

For this reason, we understand anti-surveillance fashion as a material pedagogy, insofar as it makes tangible the abstractness of AI’s constant presence as a tool of control: the garment becomes a reminder that AI is not neutral and that the body matters politically. At the same time, we do not stop pointing out the danger that it may become yet another aesthetic of the modern, since although every individual practice of resistance has political effects, its transformative potential depends on its collective articulation, which lies beyond the aesthetic intervention on one’s own body.

Conclusion: dressing as an individual and a solidaristic political act

Anti-surveillance fashion does not expect to put an end to algorithmic surveillance, but its appearance and spread signal a breaking point in which the act of dressing becomes subordinated to the existence of a control algorithm, exposing a flaw in the structure of the social system, the result of the use and abuse of an ungoverned AI.

Anti-surveillance fashion is, more than a solution to a problem, a warning that debates about the use of AI in surveillance cannot be limited to technical elements, but must open up to basic questions that remain unanswered: What options exist for people in a system that observes us permanently, and groups us without our consent?

The body becomes the last space of resistance. George Orwell already said it (2021:190–191): “We cannot act collectively. We can only pass on our knowledge from individual to individual, from generation to generation. In the face of the Thought Police there is no other way.”

References

Calvi, I. (2023). Countersurveillance aesthetic: The role of fashion in the reappropriation of identity. ZoneModa Journal, 13(1S), 75-91. https://zmj.unibo.it/article/view/17936/16931

Foucault, M. (2007). Nacimiento de la biopolítica: Curso en el Collège de France (1978-1979). Fondo de Cultura Económica.

Kronman, L. (2023). Hacking surveillance cameras, tricking AI and disputing biases: Artistic critiques of machine vision. Open Library of Humanities, 9(2), 1-35. https://olh.openlibhums.org/article/10181/galley/23437/view/

Lemke, T. (2017). Introducción a la biopolítica. Fondo de Cultura Económica.

Long, J., Jiang, T., Liu, H., Ma, C., & Yao, W. (2025). Thermally activated dual-modal adversarial clothing against AI surveillance systems. arXiv. https://arxiv.org/pdf/2511.09829v2

Monahan, T. (2015). The right to hide? Anti-surveillance camouflage and the aestheticization of resistance. Communication and Critical/Cultural Studies, 12(2), 159-178. https://publicsurveillance.com/papers/Right-to-hide.pdf

Orwell, G. (2021). 1984. Penguin Random House

Rogers, L. (2025). Styling the surveillance self in fashion media: Strategies of sexualization and sentimentality. Surveillance & Society, 23(1), 17-36. https://ojs.library.queensu.ca/index.php/surveillance-and-society/article/view/17269/12296

Treu, M., Le, T.-N., Nguyen, H. H., Yamagishi, J., & Echizen, I. (2021). Fashion-guided adversarial attack on person segmentation. arXiv. https://arxiv.org/pdf/2104.08422v2

Zhou, D., Chan, P. P. K., Wu, H., Zheng, S., Huang, R., & Zhao, Y. (2025). Physically realistic sequence-level adversarial clothing for robust human-detection evasion. arXiv. https://arxiv.org/pdf/2511.16020v2

Zhu, X., Hu, Z., Huang, S., Li, J., & Hu, X. (2021). Infrared invisible clothing: Hiding from infrared detectors at multiple angles in real world. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). https://openaccess.thecvf.com/content/CVPR2022/papers/Zhu_Infrared_Invisible_Clothing_Hiding_From_Infrared_Detectors_at_Multiple_Angles_CVPR_2022_paper.pdf

Zuboff, S. (2019). The age of surveillance capitalism: The fight for a human future at the new frontier of power. PublicAffairs.