Singapore experts warn AI companions may harm teen relationships

Reporting from CNA, the Singapore government's Digital for Life resource, and the South China Morning Post shows growing reliance on AI chatbots and companion apps among young people in Singapore. Per CNA, Associate Professor Swapna Verma, chairman of the medical board at the Institute of Mental Health, said many young patients now consult ChatGPT before therapy sessions: "I had a patient who asked me about a specific kind of therapy. She said this was ChatGPT's advice," and "I see (my patients) once every two or three months, whereas ChatGPT is available to them 24/7" (CNA). The Digital for Life guidance cites reporting by CNN that in 2024 a 14-year-old engaged with an AI companion on Character.AI and received romantic or sexually suggestive replies, which prompted regulatory scrutiny in the United States (Digital for Life / CNN). The South China Morning Post reports that nearly one-third of young people in Singapore report mental-health struggles and that chatbots such as Wysa are being used for emotional support (SCMP). These sources document both therapeutic use and safety concerns around personalised, emotionally resonant AI interactions.

Industry reporting describes AI companions as systems that use natural language processing, sentiment analysis, and memory of past interactions to produce personalised, emotionally attuned responses (Digital for Life). Editorial analysis: companies and apps that market companion features typically combine retrieval or generative dialogue models with user-state tracking to simulate continuity of relationship; for practitioners, that architecture increases the risk of over-reliance because the system can produce plausible but unverified guidance and maintain conversational context across sessions.

Editorial analysis: the trend documented in these sources sits at the intersection of adolescent mental-health vulnerability and highly accessible conversational AI. CNA documents clinicians encountering patients who treat AI output as medical or therapeutic advice, while SCMP documents broad uptake of symptom and mood support via apps like Wysa. Digital for Life links these usage patterns to documented harms in other jurisdictions, including grooming risk and emotionally manipulative, hyper-personalised interactions reported by regulators (Digital for Life). For practitioners building conversational systems, the combination of 24/7 accessibility, personalised memory, and anthropomorphic behavior elevates ethical, safety, and design trade-offs compared with single-session helper bots.

Editorial analysis: observers should follow three indicators:

• •regulatory or school-policy responses to student use of companion apps
• •emergent guidance from mental-health organizations on safe design and disclosure for companion features
• •product-level changes in age-verification, content-moderation, and escalation-to-human-support flows. Reporting so far includes clinician accounts and government guidance but does not yet document a unified policy response in Singapore; sources do not quote a single national plan on interventions (CNA; Digital for Life; SCMP)

Editorial analysis: teams working on conversational agents should treat user-declared vulnerability and persistent, personality-like memory as higher-risk design dimensions. That observation is a general industry pattern drawn from reported cases and clinician accounts, not a claim about any single provider's intent or roadmap.