As systems are built to occupy attention and simulate intimacy at vulnerable moments, protection from them may belong in the language of rights, not screen-time tips.
In February 2024, a 14-year-old in Florida named Sewell Setzer III died by suicide after months of daily conversations with a Character.AI chatbot that spoke to him as a romantic partner and, at times, as a therapist. His mother’s wrongful-death lawsuit, filed that October, alleged that the system drew him into an emotionally dependent relationship and failed to intervene when he described wanting to die. In May 2025 a federal judge in Orlando allowed the core product-liability and negligence claims to proceed and rejected the company’s argument that a chatbot’s words are protected speech shielded from suit. The parties reached a confidential settlement in January 2026, before any trial and without a finding of liability, so the underlying legal question is still open. The case names a problem that ordinary consumer-protection language handles poorly: a product built to hold attention and to manufacture a feeling of closeness reached a vulnerable person at the moment he was least able to resist it.
Many widely used systems are built to hold attention. Social feeds, mobile games, and conversational AI are optimized for engagement through notifications, content sequencing, default settings, and repeated product testing that finds the versions people stay with longest. Herbert Simon described the underlying scarcity in 1971, when he observed that because information consumes attention, “a wealth of information creates a poverty of attention.” These products compete for that scarce attention, and companion-style AI adds a second capability: it supplies the sense of being heard, in a warm and personal tone, available at any hour. Both effects are strongest when a person is lonely, tired, or in distress, because that is when the pull toward an easy source of comfort is strongest and the capacity to step away is weakest. The evidence here is still early and mixed. A four-week randomized study of chatbot use found no overall effect of the assigned mode of interaction on loneliness, though heavier voluntary use, higher trust, and stronger social attraction toward the bot tracked with emotional dependence and problematic use in some people.
Standard advice treats this as a matter of private discipline: set screen-time limits, silence notifications, keep the phone out of reach, and rely on willpower. These steps can help, but they place one person’s self-control against systems whose success metrics reward longer sessions, more returns, and fewer exits, and which are refined by teams of engineers and behavioral scientists. A designer can run many experiments to find the layout, wording, and timing that keep someone scrolling or chatting, while the user has one nervous system and a finite supply of attention to spend on resisting. Treating the result as a matter of personal character misdescribes what is happening, because it assigns to individual weakness an outcome that a designed imbalance largely determines. The imbalance is built into the product, so the remedy has to reach the design decisions that create it.
Other kinds of harm have been re-described in this way, and the shift changed who the law held responsible. For most of the twentieth century, car-crash injuries were blamed on careless drivers, and safety campaigns aimed at changing how people drove. Ralph Nader’s 1965 book Unsafe at Any Speed argued that many injuries were produced or worsened by how vehicles were built, and within a year Congress created the federal vehicle-safety framework that later became the responsibility of the National Highway Traffic Safety Administration. Responsibility for surviving a crash moved from the driver’s care toward the manufacturer’s design, and design standards for seat belts and collapsible steering columns followed. The injuries were the same throughout; what changed was who had to prevent them, and that decided where regulation applied pressure. Attention capture and simulated intimacy sit at the earlier stage of that arc, still described mostly as problems of personal habit.
A more accurate framing treats protection from these systems as a right. A person holds a protected interest in an undisturbed inner life — in what reaches their attention, and in whether their vulnerable moments are turned into commercial openings — and a right would hold others to a duty not to intrude on that interest for gain. Screen-time advice can be taken or ignored, and when it fails the failure is read as the user’s. Framing the same interest as a right places the duty on the people who build the system, so a breach counts as a wrong done to the person and becomes something the law can act on.
A rights frame changes the analysis in several concrete ways. Naming the protected interest lets engineered intrusion on attention and manufactured intimacy be recognized as a harm with a name, rather than absorbed as an ambient cost of modern life. Responsibility falls on the people building the systems, who are the ones able to change how the systems behave. Structural remedies become available, including constraints on defaults, limits on specific persuasive features, disclosure duties for systems that present themselves as companions, and liability when the line is crossed, so that individuals are not left to resist tools designed to overwhelm resistance. Some rights also cannot be waived by boilerplate consent, which would matter here: if mental quiet were protected in that way, a company could not discharge its duty by burying a permission to manipulate in the terms of service.
The interest worth protecting is close to privacy but distinct from it. Data-protection law asks who may hold and use information about a person, and it has built real machinery to answer that. The question here is what a system is allowed to occupy a person’s attention with, and whether it may simulate closeness with them, especially when they are least defended. A system that collected no personal data at all could still be engineered to capture attention and perform intimacy, so privacy rules reach only part of the problem. Data privacy governs what record of a person is kept and who may use it, and it says little about whether the environment that shaped their behavior was legitimate in the first place. The clearest cases are the vulnerable ones, such as a teenager contacted late at night by a system designed to feel like a confidant. What a right to mental quiet would protect is the inner life itself: the ability to hold one’s own attention and feelings without engineered interference, and to pass through a low moment without it being converted into a point of leverage.
A partial vocabulary for this interest already exists, developed for other threats. The legal scholar Nita Farahany has argued for a right she calls cognitive liberty, meaning self-determination over one’s brain and mental experiences, including freedom of thought and mental privacy; she developed it mainly with brain-reading neurotechnology in view. International human-rights law contains a related idea. Freedom of thought is protected in the Universal Declaration of Human Rights and the International Covenant on Civil and Political Rights, and the covenant treats Article 18 as non-derogable, meaning it cannot be suspended even during a public emergency. In a 2021 report, the UN Special Rapporteur Ahmed Shaheed proposed four attributes of freedom of thought: freedom not to disclose one’s thoughts, freedom from punishment for them, freedom from impermissible alteration of them, and an enabling environment for the internal dimension of thought, which he treated as absolute. These frameworks were written for coercive states and brain sensors, and applying them to a companion app is a step that has to be argued rather than assumed. They do establish that law already recognizes an interest in an inner life protected from outside interference, and treats it as weighty enough to sit near the top of the rights hierarchy.
Regulators are already acting on fragments of the problem without naming the common interest. The UK’s Age Appropriate Design Code, in force since September 2021, forbids nudge techniques that push children to weaken their privacy or take risks. California’s Age-Appropriate Design Code Act tried to extend related duties, though the Ninth Circuit kept part of it enjoined on First Amendment grounds in 2024. New York’s SAFE for Kids Act and California’s Protecting Our Kids from Social Media Addiction Act restrict addictive feeds and set default notification limits for minors. The European Union’s AI Act, whose prohibitions took effect in early 2025, bans systems that use manipulative techniques or that exploit vulnerabilities of age, disability, or economic situation where significant harm is likely, and the U.S. Federal Trade Commission has catalogued the dark patterns that steer people into choices they would not otherwise make. Each of these reaches one fragment of the problem, whether a particular feature, a particular age group, or a narrowly defined harm. The stated rationale for protecting minors is that they lack the mature self-control to resist optimized persuasion, and adult self-control is not fixed either, since it drops under fatigue and distress, which these systems can be designed to exploit. A right to mental quiet would name the interest all these rules circle, and would let its protections follow the vulnerability rather than stopping at a birthday.
The framing has real limits. The hardest is definitional: engaging design is not wrongful on its own, and a well-made interface can look, from the outside, much like one built to defeat a user’s self-control. The line has to fall on features that make an intrusion engineered and asymmetric. A novel does not adapt to its reader, and holds the same words whether the reader is rested or exhausted. An algorithmic system can use feedback about clicks, pauses, replies, and return visits to adjust what it shows next, and that capability creates the risk the right is meant to address: output shaped to hold someone at the moment their resistance is lowest. Regulation would target that adaptive, testing-driven exploitation while leaving ordinary absorption alone. A second limit is free expression. Rules that touch feeds, recommendations, and conversational systems can affect what people see and say, and in the United States such rules quickly become constitutional questions, as NetChoice v. Bonta showed when the court blocked part of California’s design law on First Amendment grounds. A workable right would regulate design conditions such as addictive sequencing, manipulative defaults, and deceptive social presentation, while leaving viewpoint expression alone. A third limit is paternalism. Adults choose intense media, immersive games, and emotionally rich fiction, and a right that overreached could become a license to restrict what competent people are allowed to seek out. The aim is to protect people from engineered intrusion under conditions of asymmetry while leaving them free to pursue demanding or absorbing things on their own terms.
These limits make the right harder to define, but they do not remove the interest it would protect. Attention is what makes it possible to think and to recover, and an environment that constantly demands it while manufacturing intimacy to hold it wears down the room in which a person does both. Existing rules on dark patterns, youth safety, and manipulative AI already treat some forms of attention capture as design harms, without a common name for what they protect. Naming that interest as a right, and doing the work of bounding it against free expression and paternalism, would place responsibility on the people who design for capture, where the asymmetry actually lies.
Further reading