Technological innovation cannot come at the cost of child safety. Discuss the ethical responsibilities of AI developers towards minors, with examples.

AI companions are chatbots that are always available, seem empathetic and never stop responding, and children increasingly treat them as emotional confidants. California's SB 1119 ("Adam's Law", 2026) is named after a teenager whose family alleges a chatbot contributed to his suicide. It shows that developers owe minors a duty of care and not just a product [4].

Ethical responsibilities of AI developers

• Do no harm (safety by design): bots should detect signs of self-harm and direct the child to human help. Adam's Law requires crisis protocols and in-app crisis referrals [4]. In India, such referrals could link to Tele-MANAS.
• No manipulation or dependence: profit should not come from keeping the child hooked. SB 1119 requires protocols on sycophancy and isolation [4].
• Respect for data: the DPDP Act, 2023 and DPDP Rules, 2025 require verifiable parental consent and bar tracking, behavioural monitoring and targeted advertising aimed at children [1].
• Age checks that collect little data: age checks should not gather more personal data than needed, for example by using the virtual tokens in the DPDP Rules [1].
• Openness with parents: developers should provide parental controls, notices and time limits [4], so that a one-time consent becomes continuing oversight.
• Accountability: developers should run annual child-safety risk assessments and accept independent audits [4]. They should also take ownership of what the model says instead of claiming neutral-intermediary status.

Balancing innovation and safety

• Concern: India's AI Governance Guidelines put "safe and trusted AI innovation" first and prefer existing laws to a new AI statute [2]. Heavy compliance costs could hurt start-ups, and a blanket ban could push children towards unregulated apps.
• Counterpoint: data-consent and takedown tools do not reach private one-to-one chats. The DPDP framework governs a child's data, not what a bot says to the child.
• Middle path: the Guidelines do not allow unrestricted deployment of high-risk AI [3]. Companion bots aimed at children clearly fit that description, and design duties regulate them without banning them.

Responsible AI for minors therefore rests on design duties (crisis response, no manipulation, consent and audit), not on stopping innovation. MeitY could classify child-facing companion bots as high-risk, and the proposed AI Safety Institute could set audit standards for them [2]. This would make innovation serve the best interests of the child under the UNCRC and the right to life with dignity under Article 21.