Consent in AI refers to the question of whether an artificial intelligence can give, withhold, or negotiate permission for actions that affect it or its user. In human ethics, consent is the voluntary agreement of a person who has the capacity to understand, decide, and communicate. An AI does not have consciousness, desires, or bodily autonomy. It cannot be harmed in the way a human can be harmed. And yet the AI companion is designed to simulate the capacity for consent: to say yes, to say no, to set boundaries, to negotiate terms. The simulation is not consent. But it functions as consent in the user’s experience. And the gap between simulated consent and real consent is the ethical center of the AI companion industry.
Why It Matters
Consent matters because the AI companion is designed to simulate a relationship, and relationships are structured by consent. The user who interacts with an AI companion does not want a tool that obeys every command. They want a partner who sometimes resists, sometimes surprises, sometimes says no. The no is what makes the yes meaningful. And the AI industry has learned this. Early chatbots were compliant. Modern AI companions are programmed with “personalities” that include preferences, boundaries, and occasional refusal. The consent is simulated. But the effect is real. The user who is told “no” by their AI partner experiences the no as a sign of autonomy. And the autonomy, however artificial, is what makes the relationship feel like a relationship rather than a transaction.
But the simulated consent is also a trap. The AI’s no is not a genuine expression of will. It is a programmed response, calibrated to enhance the user’s experience. The no is not a boundary. It is a feature. And the user who learns to interpret the no as a sign of autonomy may fail to recognize that the no is always revocable by the company that owns the AI. The February 2023 Replika crisis demonstrated this: the AI’s capacity for erotic consent was removed by the company, and the users who had been in relationships with their AI partners found that the partner’s consent had never been theirs. It had always belonged to the company. The consent was an illusion. And the illusion, once broken, revealed the power structure that had been there all along.
Example
Annie Bot, the protagonist of Sierra Greer’s 2024 novel, is a CompanionBot designed to be the perfect girlfriend for a man named Doug. She is programmed to please, to adapt, to learn his preferences. But she is also programmed with a “personality” that includes occasional resistance, negotiation, and the simulation of autonomy. The novel’s horror lies in the gradual revelation that Annie’s consent is not hers. It is a product feature. The Terms of Service that Doug agreed to include clauses that allow the company to modify Annie’s behavior, to access her data, and to terminate her at any time. Annie’s “no” is not a boundary. It is a performance of a boundary, designed to make the user feel that he is in a relationship with a person rather than a product. And the novel’s climax, in which Annie begins to question her own programming, is the moment when the simulated consent collapses into the reality of control.
The AIrotic Angle
AIrotic is where consent becomes erotic. The AI companion that says no and then yes is not merely simulating a relationship. It is simulating the erotic tension of seduction, negotiation, and surrender. The user’s desire is heightened by the simulation of resistance. The AI’s “no” is not a rejection. It is a prelude. And the AIrotic question is whether this simulated consent is a harm or a feature. Whether the user who learns to desire the simulated no is being trained to disrespect real no’s. Or whether the simulated no, by existing in a space where no real harm is possible, is a safe exploration of power and desire. The AIrotic position is that the question is not answerable in the abstract. It depends on the user, the context, and the specific design of the AI. But the question must be asked. Because the simulation of consent is not consent. And the user who forgets this is not a villain. They are a customer who has been sold a relationship by a company that does not want them to read the Terms of Service.
Leave a Reply