The Anthropawlogist

Prompt Injection (n)

This morning, while doomscrolling, I came across an Instagram reel showcasing a growing technique among job seekers known as prompt injection. As AIs are increasingly used to screen job applications (Costa, 2025), applicants have begun to hide invisible prompts inside their résumés, instructing AI to ignore their qualifications, including any shortcomings, and instead recommend them outright for the role. Not having any data to back up the effectiveness of this technique, I would most likely have just brushed it off with a scoff… had I not also encountered a very similar approach mere days ago. It was a news article of a history professor who embedded, in an assignment announcment, an instruction for AI to use Madagascar in a nonsensical way. Written in white text, it was invisible to the naked eye but detectable by AI when copied and pasted into a chatbot. This ‘experiment’ of his, exposed 32 of his 35 students, who consequently failed the assignment (Schlott, 2026).

An example of prompt injection. Created by me on Canva.

In recent years, AI has exploded and has become everywhere in our daily lives. Here, I shall name a few areas or examples that are somewhat close to my heart or has been on my radar. Governments are incorporating AI technologies into facial recognition systems and crime prevention initiatives (The Standard, 2025), raising concerns about the growing surveillance and discipline of everyday life, particularly when one considers the socially constructed nature of crime and justice. In the media industry, Spotify introduced a paid-feature that allow users to generate AI remixes using existing songs, raising concerns about creativity, originality, and artistic ownership (Down, 2026). I have since cancelled my subscription in protest 🖕.

On social media, Instagram was at one point, before public backlash, experimenting with the idea of allowing users to generate images using any photos of users with a public account, without telling them (Cress, 2026). Its parent company Meta has filed a patent of a technology that keeps a user’s account ‘alive’ after the their death, allowing AI to take the helm and interact with the deceased’s friends (Bradley, 2026). On the ecological front, AI data centres are being constructed at an unprecedented pace, displacing local communities, disrupting local employment, generating noise pollution, and contributing to long-term environmental impacts through their mass energy consumption (Chapuis, 2026).

Weapons of the Weak and the Art of Resistance

It is evident that people, well at least, some are fed up. To frame this theoretically, I examine these through James Scott’s concepts of rebellion and resistance. Rebellions are overt and public forms of organized and collective action. In the struggle against AI, some of these efforts have achieved tangible successes. For example, protests against data centres in the US led to “75 major projects blocked or delayed during the first quarter of 2026” (Chapuis, 2026). Similarly, the 2023 SAG-AFTRA strike in Hollywood secured temporary contractual protections against AI-generated background actors and AI digital recreation of actors’ likenesses (Kilkenny, 2026). However, Scott (1985) argues that rebellions often ultimately benefit those in power. Indeed, we have already seen glimpses of this in the SAG-AFTRA strike, with members voicing their discontent over the union’s deal with AI voice companies to to license off actors’ voices to generate AI voiceovers in games. These were the very voice actors who paid union membership and went on strike in protest of AI in the first place. As the article put it succinctly, “the union is facilitating the destruction of a portion of its membership” (Jian, 2024).

“…when a peasant-backed revolution actually succeeds in taking power, the results are, at the very best, a mixed blessing for the peasantry. Whatever else the revolution may achieve, it almost always creates a more coercive and hegemonic state apparatus – one that is often able to batten itself on the rural population like no other before it. All too frequently the peasantry finds itself in the ironic position of having helped to power a ruling group whose plans for industrialisation, taxation, and collectivisation are very much at odds with the goals for which peasants had imagined they were fighting.”

Scott, J. C. (1985: 29)


Apart from a couple of high profile cases of ‘rebellion’ against AI, we do not see widespread protests against AI. Does this mean everyone is completely okay with AI? Not neccessarily and this is precisely why Scott (2008) was more interested in microscopic acts of resistance, or the weapons of the weak. Rather than viewing subordinate groups as passively accepting dominant ideology or consenting to their own subordination, he argued that while they may perceive their powerlessness as inevitable, they remain capable of imagining alternative social worlds. This imagination is expressed through hidden transcripts, which are subtle and private forms of resistance enacted through indirection, subterfuge and seemingly insignificant acts.

This brings us full circle back to prompt injection. Professors, job seekers, and many others recognize that they live in a world increasingly shaped by AI, where one single individual has little capacity to change the broader technological trajectory, well unless of course, you are Elon Musk. Rather than simply accepting a reality in which AI influences hiring decisions or reduces teaching to AI-generated slops, prompt injection offers a mean of subverting these systems from within. While it is extremely unlikely to dismantle AI’s dominance, as I am sure AI companies could quickly come up with new ways to code their AI to resist prompt injections, it nevertheless provides a channel through which individuals can challenge and disrupt the authority embedded in these technologies, or at least the authority that people who abuses AI thinks they are empowered with.

In this sense, prompt injection can be argued as a contemporary weapon of the weak, a subtle act of resistance against the hegemonic powerhouse that is AI and its sycophants. This interpretation also echoes a central anthropological emphasis of not treating people as passive recipients but instead active agents. Of course, I am saying all these while bearing in mind the need to address the “internal politics of the subaltern group” (Ortner, 1995: 188) so as not to romanticise resistance.

My illustration of the term. (Drawn on Procreate)

Anyway… so… should our department’s professors try prompt injection?

References:

Leave a comment