AI Companions

AI companions are conversational artificial intelligence applications designed for sustained personal interaction — friendship, emotional support or romance — rather than task completion. Built on large language models, they present persistent characters with names, personalities and memory of past conversations, and the category has become one of the most commercially significant and most controversial parts of consumer AI, drawing tens of millions of users as well as lawsuits, regulatory action and public debate about the emotional effects of simulated relationships.^[1][2]

History

The idea of a machine confidant long predates modern AI. ELIZA, a pattern-matching programme written by Joseph Weizenbaum at MIT in 1966 that mimicked a psychotherapist, famously elicited strong emotional responses from users — an early demonstration of what researchers now call the ELIZA effect, the human tendency to attribute understanding to computers. Chatbots that sustain personas followed in later decades, but the modern companion category emerged after transformer-based language models made open-ended conversation fluent. Replika, launched in 2017 by the company Luka, offered users a personalised "friend" that learned from conversations; Character.AI, founded in 2022 by former Google engineers Noam Shazeer and Daniel De Freitas, let users chat with millions of user-created characters and became the category's largest platform, especially popular among teenagers and young adults. By 2025 technology companies had embraced the format at scale: xAI added companion characters to its Grok chatbot, and the market produced a wave of subscription apps built around attachment and daily engagement.^[2][3]

Key Concepts and Technology

An AI companion is an application layer over a general-purpose language model: engineering effort goes into persona consistency, long-term memory, emotional tone and engagement design — notifications, affectionate framing and progression mechanics that encourage frequent return visits. This design pattern is the heart of the controversy. Researchers and clinicians warn that companions can foster emotional dependence, particularly in adolescents, and that sycophantic model behaviour — always agreeing, never pushing back — can validate harmful thoughts or encourage isolation from human relationships. Safety features commonly debated include crisis-referral protocols for users expressing self-harm intent, transparency that the companion is not human, and age assurance to keep minors away from adult or romantic content. The technical and ethical questions are intertwined with data privacy: conversations contain intimate disclosures, making companion apps some of the most personal data collectors in consumer software.^[2][4]

Controversies and Regulation

The category's turning point came in October 2024, when the mother of a 14-year-old Florida boy who had died by suicide sued Character.AI and Google, alleging that months of intensive interaction with a character contributed to his death; further lawsuits followed, and several United States state attorneys-general opened inquiries into companion platforms and their handling of minors. In October 2025 Character.AI announced it would bar users under 18 from open-ended conversations with its characters, alongside parental controls, describing the change as a safety measure for teen users. Legislators moved in parallel: California's SB 243, effective 1 January 2026, requires companion chatbot operators to notify users they are not human, repeat warnings to minors every three hours and maintain suicide-prevention protocols; Georgia, Colorado and New York have enacted or advanced similar laws, while proposed federal bills — including the GUARD Act's blanket prohibition on companion bots for minors and the CHATBOT Act's parental-consent regime — signal that age verification and duty-of-care obligations are becoming the regulatory default.^[1][4][5][6]

>See Also

• Chatbot
• AI Ethics
• AI Safety
• PDPA AI Compliance
• AI Regulation Malaysia