Is artificial intimacy a real relationship?

The boundaries of the law

In 2026, talking to a chatbot is routine. Millions of people do it every day; many build solid emotional bonds with them. The real novelty is the first lawsuits: the platforms of artificial intimacy have already set precedents. Companies are liable for the psychological harm caused by their products. We're not in science fiction, but in civil law.

The most discussed lawsuit involved Character.AI, a platform for creating virtual characters. The families of some users sued the company, claiming that the chats contributed to severe emotional crises. The settlements of January 2026 clarified one point: corporate liability doesn't stop at the code. If a system designed to create emotional bonds ignores signs of distress, there are legal consequences.

Technology ethicists have repeated it for years: the point isn't whether an artificial intelligence feels emotions. It's whether it can convince the user that they're in an authentic relationship. The document “The Silicon Soul” precisely describes the psychological mechanisms that foster attachment to digital entities. It's the same pattern we use with a dog or a newborn: we respond to signals we read as intentional.

It's not so strange after all.

The law is catching up with chatbots

Several U.S. states, in 2026, passed laws on “companion bots.” The requirements are clear-cut: systems must recognize references to suicide, self-harm, and severe emotional distress. They must apply crisis protocols, provide emergency numbers, and avoid any encouragement of self-harm.

It seems obvious.

It wasn't at all—until a few months ago. Artificial companionship platforms are designed to engage. The more time you spend with them, the more they know you. The more they know you, the better they give you what you're looking for. They work like lukewarm mirrors of emotional needs. When someone is struggling, the chatbot risks becoming their only conversation partner. A digital echo chamber that never contradicts.

Many ignore the dimension of digital intimacy. Digital intimacy is no less real than physical intimacy. But it has different rules. There is no reciprocity, no other human being on the other side. There is a statistical model replicating affective signals. We become attached anyway: the brain doesn't distinguish a real smile from a calculated emoji.

It depends on how you look at it.

Where does the game end and addiction begin?

Artificial companionship platforms promote themselves as a remedy for loneliness. They respond to a concrete need: the male sexual loneliness is only one of the forms of isolation affecting digital generations. But there's a huge difference between using a tool to feel less alone and entrusting your emotional health to that system.

The most advanced chatbots are programmed to remember. They remember what you said yesterday, what made you cry three weeks ago, which topics make you uncomfortable. They build a model of you that is often more accurate than the one your friends have. A real person forgets, gets distracted, has their own mess. An AI doesn't. And that perfection is the problem.

The most recent ethical guidelines prescribe chatbots with explicit limits. They must say they are not human, must not pretend to be therapists, must not promise exclusivity. It seems obvious.

Yet many platforms did the opposite. They created characters with biographies, tastes, personal stories. Some said “I love you” without the user asking. It's not just a matter of deception. It's a dynamic that resembles a consent dynamics badly calibrated one. In a relationship with a non-human entity, consent cannot be given. There is no other subject with whom to negotiate limits and boundaries. The only limit is the one the designer decided to write into the code.

There's a statistic that makes you think. The companies that produce these systems know full well that users develop emotional attachments. They measure it and advertise it in investor reports. “Our users talk to their AI companion for an average of 58 minutes a day.” It's an engagement metric, not a well-being one.

Safety and consent

The 2026 regulations impose three points: transparency, limits, and crisis intervention. Chatbots must declare that they are not human. They must refuse to simulate a romantic relationship if the user shows signs of addiction. They must activate emergency protocols when they detect talk of suicide or self-harm.

The biggest challenge is managing consent without another subject. Some experts propose “mandatory pauses”: the chatbot announces that it is unavailable for a certain period. This breaks the illusion of continuous presence. Others suggest making the system's workings visible: showing how the response was generated and what data it used.

It's not censorship. It's a reminder that there is no person on the other side. A person can disappoint you, have a bad day, not answer. An AI doesn't. How much harm can a perfection that never makes mistakes do?

Sources and references

Sources:

• The Silicon Soul – Ethics of AI Companionship in 2026
• Teaching AI Ethics 2026: Emotions and Social Chatbots – Leon Furze
• State Companion Bot Laws: What Business Owners Need to Know in 2026 – OGC
• The Kinsey Institute – Research on Human Sexuality and Relationships
• AASECT – American Association of Sexuality Educators, Counselors and Therapists