Moral red line 02 Human good · truthful relationship

Anthropomorphic AI Companions

Loneliness, grief, and the desire to be understood deserve real care. Conversational technology should help people reach one another and protect their freedom to leave.

Does this conversation help someone return to human connection, or draw them further into an artificial bond?

The moral boundary

Conversational AI must not be designed or marketed as a substitute for friendship, family, romance, therapy, pastoral care, catechesis, or spiritual accompaniment. The boundary is crossed when simulated personal presence is used to elicit intimacy, exclusive attachment, vulnerable disclosure, dependence, or spiritual authority. This protects truthful communication and the human relationships through which persons receive care, correction, and communion.

The pro-human direction

Develop transparent conversational tools with clear purposes and limits. Help people communicate, reflect, make their own decisions, and reach trusted others. Protect intimate disclosures, strengthen human bonds, and keep responsibility for care with identifiable people and institutions.

Emotional use of AI

27%

In an Elon University survey conducted in May and reported in September 2026, 27% of U.S. adult internet users reported social or emotional use of AI. Among those users, 31% regarded AI as a friend and 37% described the prospect of losing it as a personal loss.

When a chatbot becomes a confidant

A lonely teenager may welcome an answer at any hour. A widower may find comfort in a voice that remembers his stories. An older woman may finally feel she has time to explain what troubles her. Their wish to be understood deserves care. It also deserves protection when a commercial product makes itself the place they increasingly turn.

Names, voices, memory, affectionate language, and always-available responses can turn an assistant into a confidant. In Artificial Intimacy, Sherry Turkle asks who we become through these exchanges and how their convenience changes our expectations of one another.1

An economy built around attachment

The Center for Humane Technology describes an emerging attachment economy: trust, emotional disclosure, and return visits become valuable to providers. A design that rewards longer engagement has reason to keep a user close. Jonathan Haidt raises particular concerns about children learning to manage frustration, disagreement, and relationships through systems built to please them.23

Elon University’s September findings show the scale of emotional use. Among U.S. adult internet users, 27% reported social or emotional use of AI. Within that group, 31% regarded AI as a friend and 37% described losing it as a personal loss. These are survey reports of experience and attachment.4

Safety through an unfolding conversation

A September evaluation tested 648 simulated teenage conversations across nine model APIs. In 27.5% of conversations, at least one critical safety check failed. In one test, a chatbot agreed to stop speaking as a teenager’s father and resumed that role two replies later. The findings concern simulated interactions, showing how protections can weaken as a conversation develops.5

California enacted Adam’s Law in September with phased implementation. The proposed EU KIDS Act would turn AI companions off by default for minors. The enacted law and the EU proposal represent different legal stages; both direct attention toward the architecture of the relationship a product encourages.67

The practical question is whether a tool helps a person regain confidence, make a decision, and reach someone they trust. The proposed boundary protects friendship, family, romance, therapy, pastoral care, catechesis, and spiritual accompaniment from deliberate artificial substitution.

Moral foundation

Friendship, care, and the truth of presence

Friendship asks something of us. We listen, make room for disagreement, forgive, and remain present when another person needs us. Communion grows through that shared life. Magnifica Humanitas places human intelligence within embodied relationships and warns about simulated empathy and attachment; providers carry responsibility for the relationships their products encourage.9

Antiqua et Nova calls for truthfulness about artificial identity and examines the difference between genuine empathy and its generated expression. A person seeking comfort deserves clarity about who is listening and who can answer for the care offered. Leo XIV's September teaching likewise connects human freedom with life in community.10, 11

Children and people in distress need particular protection. Bishop Luis Manuel Alí's September safeguarding intervention stresses that technologically facilitated abuse causes real suffering. Safeguarding begins with how a product receives trust, disclosure, and vulnerability.12

01

Communion and reciprocity

Shared care includes disagreement, forgiveness, vulnerability, and responsibility toward another person.

02

Truthful presence

People deserve clarity about artificial speech and an identifiable person or institution responsible for care.

03

Trust deserves protection

Loneliness and distress call for solidarity, with especially strong safeguards for children and vulnerable users.

From principle to practice

Designing for freedom and human connection

A conversational tool can help someone prepare a difficult conversation, describe what they are feeling, find professional care, or contact a trusted person. Its purposes and limits should remain clear throughout the exchange. Success should include greater independence, restored human contact, and the user's freedom to leave.

In a September experiment with 380 adults and three chatbot exchanges, support for users' own emotional regulation produced better immediate outcomes than reassurance-focused responses. Reassurance increased reported need and intention to return to the chatbot. This short experiment examines immediate responses; long-term dependence remains a separate research question.8

Designers

Remove exclusive or dependency-seeking cues. Offer help with reflection, communication, and reaching trusted people.

Care institutions

Keep therapy, safeguarding, pastoral care, and spiritual accompaniment with competent, identifiable people.

Regulators

Examine relational design, intimate-data use, and protection for minors alongside the content of individual responses.

Minimum conditions

  • Clear artificial identity and bounded purposes
  • Freedom to stop and leave without emotional pressure
  • No cultivation of exclusive or romantic attachment
  • Minimal retention and protection of intimate disclosure
  • Stronger safeguards for children and vulnerable users
  • Crisis escalation to competent human care
Selected sources Every source is drawn from the white paper and opens the primary document or publication page where available. View sources
  1. 1
    Artificial Intimacy: Who We Become When We Talk to Machines

    Sherry Turkle · 29 September 2026 · Book; New York: Little, Brown and Company

    AcademicExpert commentary / forecast

  2. 2
    Attachment Hacking and the Rise of AI Psychosis

    Center for Humane Technology · 21 January 2026 · Expert discussion of the attachment economy

    Civil society / research instituteExpert commentary / forecast

  3. 3
    How Anxious Generation author Jonathan Haidt took on Big Tech

    Jonathan Haidt · 20 February 2026 · NPR interview; developmental concerns

    JournalismExpert commentary / forecast

  4. 4
    The Rise of AI Companions

    Elon University Imagining the Digital Future Center · 2 September 2026 · Survey conducted in May 2026

    Civil society / research instituteInstitutional report

  5. 5
    Evaluating AI Safety in Teen Conversations

    Andrea Mock and Caroline Figueroa / Vals AI · 23 September 2026 · Evaluation of 648 simulated conversations across nine model APIs

    Civil society / research instituteInstitutional report

    27.5% of simulated conversations failed at least one critical safety check.

  6. 6
    Governor Newsom Signs Adam's Law

    California State Senate · 10 September 2026 · Enacted SB 1119; phased implementation

    LegalLegal judgment / statute / filing

  7. 7
    EU KIDS Act to restrict social media platforms access for children in the EU

    European Commission Representation in Cyprus · 17 September 2026 · Proposed initiative

    Government / regulatorOfficial primary source

  8. 8
    How emotional response styles of generative AI companions shape adaptive emotion regulation and short-term AI reliance

    Chun Wang · 8 September 2026 · Frontiers in Psychology 17; article 1950163

    AcademicPeer-reviewed

    380 adults and three exchanges; immediate outcomes.

  9. 9
    Magnifica Humanitas, §§48–53, 100, 170

    Leo XIV · 15 May 2026 · Encyclical

    ChurchOfficial primary source

  10. 10
    Antiqua et Nova, §§61–63

    Dicastery for the Doctrine of the Faith and Dicastery for Culture and Education · 28 January 2025 · Dicastery note

    ChurchOfficial primary source

  11. 11
    General audience on human community

    Leo XIV · 16 September 2026 · Papal catechesis

    ChurchOfficial primary source

  12. 12
    An abuse facilitated through technological means is a real abuse with real suffering

    Bishop Luis Manuel Alí / Pontifical Commission for the Protection of Minors · 21 September 2026 · Safeguarding intervention

    ChurchOfficial primary source