The Self You’re Protecting Doesn’t Exist

Sartre, bad faith, and what AI really exposes about authenticity.

Much of what I have written about AI and authenticity thus far has cautioned that over-reliance on the former can erode the latter. Specifically with respect to capacities for critical thinking, and self-awareness. This position rests, almost entirely, on an assumption Sartre spent his career dismantling which is that there’s a fixed, genuine self in there to be authentic to. This essay pulls that assumption out, resulting in a stranger and more interesting debate.

Sartre’s starting point is the famous inversion: existence precedes essence. For objects such as a hammer or a table, essence comes first. The craftsman has a concept of the thing before making it. However, humans, Sartre argues in Being and Nothingness, arrive first and define themselves afterward, through their choices and actions. There’s no blueprint. No inner core waiting to be discovered or expressed.

This is what he means when he says we are “condemned to be free.” Not free in the feel-good sense. Free in the vertiginous, no-exit sense wherein we can’t appeal to a fixed nature to justify what we do, because there isn’t one. Every excuse along the lines of “that’s just who I am” or “I had no choice” is, in Sartre’s framing, a form of bad faith: mauvaise foi, the self-deception of pretending you’re a thing with fixed properties rather than a consciousness that keeps choosing itself into existence.

His famous example is the waiter who performs his waiter-ness with such perfect precision – the too-rigid posture, the slightly-too-attentive manner – that he seems to be being a waiter the way a stone is a stone. He’s fleeing from his freedom into a role. Lying to himself. That’s bad faith.

Relating this to AI, the conventional critique of AI-assisted writing is that it’s inauthentic, it’s not really you. Sartre’s framework immediately complicates who that “you” even is. If the self is not a fixed thing but a continuous act of self-creation through choices, then authenticity can’t mean fidelity to some stable interior voice. It has to mean something else. What it means, for Sartre, is owning your freedom. Acknowledging that you chose, and that you are responsible for the choice. Authenticity isn’t about the content of what you produce, it’s about whether you’re honest with yourself about the fact that you’re producing it, and why.

Which opens up a genuinely uncomfortable question: what does that look like with AI? You can use AI in bad faith. Probably the most common version of that is “everyone does it, I had no real option” rationale treating the tool as something that happened to you rather than something you reached for. That conceptualization is a way of dissolving responsibility into social pressure, which is exactly the kind of self-deception Sartre diagnoses. You chose. The question is whether you’re acknowledging that.

A less obvious point is that you can also be anxious about AI in bad faith. The person who insists their writing must be entirely unassisted, who treats that commitment as a fixed feature of their identity rather than a live choice they keep making. That’s its own kind of rigidity. Sartre is suspicious of any position that functions as an escape from the ongoing burden of choosing. Even virtuous-seeming ones.

Yet another angle worth consideration is how Sartre distinguishes between two modes of being: pour-soi (for-itself) and en-soi (in-itself). Human consciousness is pour-soi: it has no fixed nature, it’s always ahead of itself, perpetually becoming. The in-itself is the mode of things: complete, determined, without interiority.

A language model is, in this framework, definitively in-itself. Not because it lacks sophistication which it clearly doesn’t, but because it has no freedom, no self-deception, no anguish about its choices. It cannot be in bad faith. It can’t be authentic either. Those categories don’t apply to it any more than they apply to a very elaborate hammer.

Which means the authenticity question has never really been about the AI. It’s about the human holding it. The tool doesn’t make the choice bad faith, the self-deception does. And self-deception was available long before large language models. Ghost-writing, research assistants, editorial feedback that rewrites half your sentences. Humans have always used other minds to produce work attributed to themselves. The question of what we’re honest about with ourselves, and with our readers, is old.

Perhaps what AI does is make the question more pressurized and harder to dodge. The assistance is so complete, so capable of replacing rather than supplementing, that the old ambiguities get harder to sustain. You can’t quite tell yourself the same comfortable stories. The deeper Sartrean provocation is this: the widespread anxiety about AI and authenticity might itself be a form of bad faith. Not because the concern is wrong, but because of how it’s often held as if there were a precious, pre-given authentic self at risk of contamination, something that existed fully-formed before you started writing and that writing is supposed to express.

That picture is exactly what Sartre wants you to give up. The self isn’t prior to expression. It’s constituted through it, through the accumulating weight of choices. Losing the comforting idea of an inner true self doesn’t make authenticity meaningless, it makes it harder and more demanding. You can’t just be authentic by virtue of what you are. You have to keep choosing it, in full awareness that you could choose otherwise.

That’s an uncomfortable place to stand. Which is, probably, why most people prefer the simpler story about AI faking the genuine article.