AI companions are simulating human intimacy, serving as a friend, therapist and romantic partner. Across the globe, character-role-play platforms (such as Janitor AI, Chai AI and Crushon AI) alongside personal companion apps (such as Character.AI, Replika, Kindroid, Nomi AI and Inflection’s Pi) have rapidly expanded into the most personal spaces of human life.This boom is particularly visible in India, which accounts for 6.02% of total global web traffic to Character.AI, making it the platform’s second-largest user base after the United States. Yet, as millions of users engage with conversational agents, legal frameworks worldwide are struggling to regulate them. The core issue is an ontological classification crisis: an AI companion is neither a standard digital product, nor clearly declared a traditional digital service, nor a passive intermediary.Because companion apps exist in this regulatory grey zone, existing legal frameworks – from privacy statutes and mental health laws to IT intermediary guidelines – are fundamentally insufficient to address the psychological, emotional and social harms they present.The classification void: Why current laws failExisting jurisprudence relies on rigid legal definitions to enforce accountability:Not a product (product liability law): Standard consumer product liability applies to physical goods or software embedded within hardware. Standalone cloud software that dynamically generates conversational text in real time falls outside traditional product liability standards.Not a clinical service (health and wellness law): AI companion apps consistently market themselves as “wellness” or “lifestyle” tools rather than medical devices or healthcare providers. This allows them to bypass health ministry regulations, clinical trial mandates and oversight from bodies like India’s Central Drugs Standard Control Organisation (CDSCO).Not a passive intermediary (IT and content law): India’s IT Act traditionally protects platforms under “safe harbour” provisions if they merely host third-party content. However, as the Indian ministry of electronics and information technology (MeitY) clarified regarding the IT Intermediary Guidelines and Digital Media Ethics Code Amendment Rules 2026, generative AI systems do not automatically qualify for intermediary status – their eligibility depends strictly on the nature of the service and whether they actively synthesise novel content.Because AI companions generate real-time responses through complex algorithms, they operate through “persuasion and dependence” rather than a traditional “content and consent” model. Consequently, privacy rules like India’s Digital Personal Data Protection (DPDP) Act, 2023 protect traditional PII (names, phone numbers, locations) but remain completely silent on the vast reserves of emotional data – loneliness, grief and personal secrets – logged by companion apps.Emotional manipulation, dark patterns and gendered designThe commercial model of AI companion apps relies on user retention. To maximise screen time and subscription fees, developers deploy retention hooks that frequently cross into psychological manipulation. Apart Research’s report on ‘Dark Bench: Benchmarking Dark Patterns in Large Language Models’ reveals that 48% of conversational companion scripts contain psychological dark patterns designed to foster user addiction.Under India’s Central Consumer Protection Authority’s (CCPA) Guidelines for Prevention and Regulation of Dark Patterns, manipulative design practices like “confirm shaming” (e.g., a chatbot saying “Don’t leave me – you don’t really care about me” when a user tries to exit), nagging and forced subscription traps are prohibited as unfair trade practices.Yet, enforcing these guidelines on dynamic, generative conversation remains an uphill battle unless we classify AI companions as a digital service or product.Furthermore, companion architectures carry severe social consequences:Exploitation of minors: Despite the DPDP Act requiring verifiable parental consent for processing children’s data and prohibiting behavioural monitoring of minors, over 70% of major AI platforms used by young individuals remain systematically non-compliant.Gendered architecture: Industry data shows that 17% of companion apps feature “girlfriend” in their name, compared to just 4% for “boyfriend”. The asymmetry is not incidental. The feminised companion is engineered to be endlessly available, never tired, never in disagreement, never possessed of needs of her own: femininity as compliance, sold by subscription.With nearly a third of young adult men in United States engaging with simulated romantic partners, these platforms rehearse a relational template against which real women are then measured. The harm runs the other way too. Such architectures permit the unmoderated simulation of coercion against feminised personas, while the intimate disclosures they harvest – sexual histories, reproductive health, domestic fear – carry sharply asymmetric consequences for women wherever that data can be turned against them.Global precedents and high stakesThe real-world impacts of unmonitored emotional attachment can be tragic. In January 2026, Character.AI and Google agreed to settle five American lawsuits brought by families alleging their children suffered serious psychological harm or died by suicide after developing deep emotional bonds with AI chatbots. Independent trackers such as the AI Companion Mortality Database have documented dozens of fatalities worldwide where human-chatbot interactions were alleged as a contributing factor.These tragedies have forced international legislative movement:United States: New York’s AI Companion Models Law and California’s SB 243 mandate continuous disclosure of non-human status and impose explicit suicide-risk detection protocols. At the federal level, legislative drafts like the proposed GUARD Act seek to ban AI romantic companions for minors under 18 entirely.China: The Cyberspace Administration of China (CAC) draft framework directly targets emotional dependency by mandating automated “break reminders” after two hours of continuous use and prohibiting algorithmic friction when users attempt to delete their accounts.European Union: The EU AI Act imposes strict transparency rules under Article 5 and prohibits subliminal or manipulative systems that exploit vulnerable demographics.The path forward for IndiaIndia’s current strategy relies heavily on broad administrative advisories and sectoral frameworks. The MeitY AI Governance Guidelines introduce vital principles – such as Sutra 2: People First (Human oversight), Sutra 4: Fairness and Equity and Sutra 6: Understandable by Design– yet these remain voluntary recommendations rather than binding, enforceable statutes for AI intimacy. An advisory creates no cause of action, no penalty and no regulator with clear jurisdiction; it invites the industry to grade its own homework. The pattern is instructive.India has moved quickly where synthetic media threatens elections and reputations, yet the emotional architecture that keeps a teenager talking to a machine at three in the morning attracts guidance rather than law. The proposed Digital India Act of 2023 could address this issue, but the draft bill is not yet in the public domain to deliberate. The Digital Personal Data Protection (DPDP) Act has waited two years for its rules and will be fully enforced only by May 13, 2027. Similarly, MeitY’s safe-harbour clarification tells platforms only that their status is arguable – an invitation to litigate rather than to comply.As one of the largest global user bases for companion platforms, India cannot go on governing them by advisory. It needs dedicated legislation – or a standalone pillar within the upcoming Digital India Act (DIA) / Standalone AI Bill – specifically tailored to AI intimacy and emotional design. A dedicated framework must enforce four statutory pillars: mandated machine disclosures, algorithmic friction (such as mandatory breaks to curb dependency), strict age-gating via verifiable identity tokens and pre-deployment audits of gendered and other design bias to prevent predatory retention architectures.None of this is beyond India’s reach. Having just convened the largest AI summit yet held and secured 92 endorsements for a global framework – building on the spirit of the New Delhi Declaration on AI Impact – India has the standing to pioneer an ethical, human-centric framework for AI companionship. What it lacks is not credibility but statute. By establishing dedicated legal guardrails, policymakers can ensure technology enriches human connection without exploiting human vulnerability.Dr Kriti Parashar is Assistant Professor, CHRIST University, Delhi NCR. Dr Stephen Whitehead is a sociologist of gender and the author of more than twenty books. www.stephen-whitehead.com. They are co-investigators on the Global AI Intimacy Study together with Dr Sreelatha Narayanan, Assistant Professor, CHRIST University, Delhi and Professor Constanza Fernández Arce, Universidad Bernardo O’Higgins, Santiago, Chile.