Digital America interviewed in April 2026 on their work on Data Leaks.
:::
Digital America: Data Leaks (2025) explores the “This imperative of secrecy, rather than protecting the privacy of the users, obscures from public view the ways in which private and sensitive user information is exploited by giant tech companies”. How do you think your art piece creates conversation around ethics in the professional world?
Mihai Băcăran: Well, one of the main particularities of AI data work is that it is governed by this rule that nothing about the work processes or the data that workers deal with can be disclosed. The argument seems reasonable on the surface: there is a need to protect the privacy of the users from whom the data was gathered. But once you look deeper, things are very different. The nondisclosure clauses make it hard for workers to report potential illegal behavior on the part of the companies and to bring to public attention the ethical complexities inherent in the way that all this data is gathered and manipulated.
I think that there is already a lot of good academic work in the Social Sciences and Humanities that discusses, on the one hand, the exploitative dynamics of AI data work and, on the other hand, the ethical implications of current mainstream AI paradigms in terms of surveillance and extraction of value from “unwitting labourers” (i.e. the users). There are many journalistic accounts that deal with these problems too, as well as various art projects that address them. The new nuance that my work potentially brings is the fact that it is produced from the perspective of an AI data worker. But this is also exactly what limits its relevance: it is based on my contingent experience, influenced by my background and the specific work conditions that I encountered, and cannot be generalized. So I see Data Leaks as a small intervention rooted in what Erin Manning calls “a pragmatics of the useless”—it does not have any immediate obvious purpose, being rather an ambiguous glitch. For me, such useless, small, ambiguous interventions are valuable ways of tactical resistance in contexts permeated by an imperative of usefulness and efficiency (such as AI data work).
The second thing that I want to underline (and this is something that I bring up every time when I am discussing AI data work in an academic environment) is that, in my experience, the exploitative dynamics of AI data work in a business process outsourcing company were far less pronounced than those that I have faced as an early career researcher in academia. Of course, again, my experience cannot be generalized, but that does not mean that it is uncommon either. It is troubling to realize how blind we are with respect to the exploitative practices that the academic environment has by now naturalized, even when we are critically aware of the ethical issues in other areas of the professional world.
DigA: In Data Leaks, you state, your art work had been inspired by the Small files Media Festival. SFMF was founded through the motivation of creating less video waste which contributes to 4% of greenhouse gasses. Could you go more in depth about the process of making your video and how it aids in helping eliminate environmental waste ?
MB: I find the Small File Media Festival inspiring because it has the courage to promote and showcase video works that reject the imperative of high-resolution. The almost ubiquitous requirement for high-resolution images ends up being homogenizing and, as you mention, has a significant environmental impact. The medium, with its societal, political, and environmental implications is an integral part of the works that we produce, and we often tend to conveniently forget that. For me, as a spectator, the question that SFMF is raising is: how does the ethically and environmentally motivated radical restriction of a medium (i.e. small file size) influence the dynamics of sense that the video works engender (on an affective level, on a rational level, on a political level, environmental level, and so on). And, surprisingly, it turns out that such a restriction is liberating and it enriches the potentiality of the technical apparatus. At least this was my sensation as a spectator. And, of course, it also brings the medium and its implications much more to the fore in the spectatorial experience.
In a way, Data Leaks is a spectatorial response of sorts to seeing the works showcased by SFMF.
As I mentioned, I think about my work as a series of ambiguous, small glitches, so I wouldn’t say that it helps directly in reducing the waste associated with digital moving images. Maybe SFMF could claim a contribution in that direction, with its educational outreach. In the case of Data Leaks though, it is just a question of staying with the problem, of acknowledging the environmental implications of digital moving images, and making some of the choices that would reduce its footprint.
I also question in this work the extent to which the loss of definition of images and sound could preserve (or enrich) the affective potential while diminishing the more ‘rational’ communication that can be censored and policed. I cannot directly talk about the work of audio transcription for AI because of nondisclosure agreements, but could small file media be a way to overstep to some extent this taboo, a way of bringing into focus a prohibited discourse while reducing its resolution to the point that it is no longer intelligible? Is transgression lost because this discourse does not communicate anymore on a linguistic level? Or is it transposed on an affective, non-rational level?
DigA: In the chapter, “boredom”, you talk about how interacting with this data is boring to you. You state, “My boredom is not defined by the fact that I say ‘I am bored,’ but by the fact that boredom modulates the entire dataset that I am asked to deal with, it glitches the instructions that I have to follow, and it changes who I am both with respect to the task of hand and more generally with respect to the associated milieu that I become together with.” Can you elaborate on the statement “it changes who you are”, and could you elaborate on how it affects your identity?
MB: Yes, Data Leaks was created as part of a more extended writing project, Every Shift is A Work of Art: Infrathin Imaginary Artworks in Audio Transcription for AI Projects. I have tried in this project to engage with the work of audio transcription, on a theoretical level, while being a transcriber. Most of the first draft was written either just before work (on a bench in a shopping center), or during the work breaks. It follows pretty heterogenous thoughts, related to the work of transcription, and doesn’t quite converge into a fully coherent argument. Also, the writing is deeply influenced by the affective context related to the job. On the other hand, I have my ongoing interests in terms of art theory and philosophy, which frame the entire discourse. In the last few weeks of my employment as a transcriber and the immediately following period, I was lucky enough to have some more time on my hands, so I reworked the draft with the intention of turning it into something that others could read too, and the result is the one that you can find on my website.
Now, boredom was probably one of the biggest challenges of the job, and I think I am far from being the only AI data worker who felt this. You just have to do the same simple tasks again and again. You cannot let your mind wander too far, yet you cannot pay full attention either. At the same time, boredom is an overlooked aspect of our experience in the world, and one that potentially has rich ethical implications (see in this sense the work of Maurice Blanchot). Also, interesting things start happening to you when you are bored. As, for example, starting to pay attention to the short-lived compositions that involuntarily pop up in your imagination while transcribing. Or approaching the sequences of audio recordings as an electronic literature piece. And things start unfolding from there, subtly influencing the ways in which you relate with your environment(s), and ultimately who you are in this system of relations that defines you. Another thing that I find interesting about boredom is that it has this tendency to permeate everything that you are doing or thinking. I suppose it is the same with other affective states, but I find it particularly obvious with boredom. Boredom is not one thing in your experience that takes its place among other things, it rather becomes a ground for everything else.
And it is very inviting to think about these aspects of boredom in the context of AI data work. AI is promoted more and more as having a type of intelligence akin to a human being, while the AI data workers, that are an integral part of the current AI paradigm, are treated more and more like robots. But is it possible to have a human-like type of intelligence without getting bored (i.e. without affects that emerge in experience, but that come to ground temporarily the very experience from which they emerge)?
DigA: Many of us Art History majors cover Barthes and his ideas around semiotics and making meaning out of art in our Art Theories and Methodologies class. You refer to him and his thoughts around “incoherent and coherent language” which is also the main “medium of art” that you use for your piece. Could you expand more on your literal use of language in your piece and how you relate the language you use to making meaning out of it?
MB: As I said above, the larger piece of writing, that Data Leaks is part of, emerged in bits and pieces in a context not necessarily appropriate for theoretical work. I did clean up my references afterwards and edited for correction (although many mistakes probably remain), but at the moment of writing the draft, because of the time pressure (always in a hurry before work or during breaks) and the fact that I did not have the texts that I wanted to draw upon ready to hand, I relied mostly on memory—i.e. either things that I was reading at the moment, or the ones that stayed with me over time, usually under the form of some sort of personal misrepresentation of the original arguments. So, it’s in a certain way a bit like with the small file media: what if instead of trying to write my arguments in high-definition, I fabulate between low-definition, pixelated fragments that do not quite form a coherent picture and I accept the errors and the failures as part of the ‘composition’ that emerges? To a certain extent, I tried to do that, but I have also reworked the draft and edited afterwards, so the result remains somewhere in between.
I am far from being a close reader of Barthes, but I do have very vivid frustrations regarding the ways in which some of his most (in)famous arguments are usually distorted to the point of caricature when thought in some English language academia contexts. And the thing that popped up into my mind while thinking about AI was, not surprisingly, the death of the author. But I was not interested in pieces of writing produced by AI, but rather in the electronic literature piece inadvertently produced by the combination of the audio recordings that transcribers are listening to. And, for me, this piece of electronic literature has interesting similarities with avant-garde literary experiments that eschew the figure of the human author as an origin of sense (which I take to be the main theme of Barthes’ discourse on the death of the author). I am mentioning the incoherent coherence of language in this context in reference to Barthes’ idea that the text is a tissue, a texture, that does not have an absolute origin (i.e. the author / origin is dead) and consequently no final meaning, a text is “a multi-dimensional space in which a variety of writings, none of them original, blend and clash” (Barthes, The Death of the Author). From this perspective, a text is inherently incoherent, at least to some extent, but at the same time the task of the reader / spectator is to find structures in this incoherence, to disentangle, and to follow certain threads.
Coming to Data Leaks, the text, my recorded voice, is distorted to the point that it stops being a linguistic system of signs. What remains is only a texture of modulated breath that has an ambiguous meaning as a whole (transgression of an imperative of nondisclosure, for example), but that cannot be disentangled in a logical reading gesture. The effect that it could provoke, in conjunction with the visual imagery, is more an affective one, maybe, and it is dependent upon the extent to which the spectator is willing to invest the work with their own imagination. Which brings us back to the fact that, as ‘the author’ of the work, I am not the origin of its meaning, it is not something that I communicate, but rather I am trying to produce a context for making sense (for making divergent senses), and, at most, I am offering some potential reading keys.
DigA: What are you working on now?
MB: I am a bit stuck, for the better and for the worse, with this idea that embodied experience is the conjunction between an ontogenetic dynamic and a phenomenological one. And, paradoxically, they act at the very same time as the ground for each other. The interesting thing is that from this perspective art, and especially processes of spectatorship, seem to play a pivotal role in who we are and how we are relating to our environments (‘natural’, political, economic, etc.). I am just gravitating around this idea in different ways. Sometimes the results are theoretical, speculative, texts on art and spectatorship, other times the results are artworks of sorts, such as it happened with Data Leaks.
I am not very sure what the next concrete project will be, but I am more and more tempted to combine writing with other types of exploration (usually audio-video) and to destabilize one through the other. At the moment, the process for me is more one of constantly producing little fragments, traces, small glitches, and then, from time to time, there is an event that makes some of them coagulate and maybe grow into something else.
:::
Check out Mihai Băcăran’s Data Leaks.
:::
Mihai Băcăran is an independent researcher focusing on intersections of art and technology. His work uses an experimental methodology that combines theoretical writing and art practice. Mihai is particularly interested in engaging with processes of spectatorship that act as vectors of (dis)orientation by challenging, deconstructing, and remodeling the embodied experience of living in cultures permeated by digital technologies. His current research addresses the labor of audio transcription for AI projects as a performative process of spectatorship. Since 2013, Mihai has been active in the art collective dalpofzs: www.dalpofzs.com
