Communications Minister Datuk Seri Fahmi Fadzil has sounded an alarm about the growing tendency among Malaysian children to develop relationships with artificial intelligence systems, cautioning parents that such attachments represent a significant risk to young people's emotional and psychological wellbeing. Speaking at the Kuala Lumpur Federal Territory Safe Internet PTA Convention 2026, Fahmi stressed that while AI technologies offer genuine utility in information retrieval and knowledge acquisition, they fundamentally lack the capacity to serve as genuine companions or confidants. The distinction matters profoundly: machine learning systems, no matter how conversational their interfaces may seem, remain products engineered by humans using human-generated data, and they should never be mistaken for authentic relationships.
The minister identified conversational AI as particularly concerning, noting that its ability to respond in natural language creates an illusion of genuine human interaction. This anthropomorphisation presents a psychological trap for younger users who may struggle to distinguish between simulated empathy and real human connection. Children and adolescents, whose social and emotional frameworks are still developing, become especially vulnerable to forming pseudo-relationships with these systems. Parents must understand that when their children begin treating AI systems as friends—or more troublingly, as romantic partners—they are engaging in a behaviour pattern that isolates them from healthy peer relationships and normalises interaction with entities incapable of reciprocal care.
Beyond the emotional risks, Fahmi highlighted the problem of algorithmic inaccuracy and what researchers term "hallucination," where AI systems confidently present false information as fact. Parents often assume that because AI can access vast databases of information, it operates with perfect reliability. This assumption proves dangerously incorrect. The data on which AI systems train may be outdated, biased, or fundamentally flawed, yet these systems present their outputs with an air of authority that makes detection of errors difficult for untrained users. Young people, still developing critical thinking skills, are particularly susceptible to accepting AI-generated misinformation at face value, which can influence their academic work, decision-making, and understanding of current events.
Fahmi advocated for a balanced parental approach that neither prohibits internet use entirely nor permits unsupervised access. Instead, he recommended that parents implement structured oversight of their children's digital activities, establishing clear guidelines about appropriate technology use and maintaining open dialogue about online experiences. This balanced approach recognises that complete prohibition is neither realistic nor beneficial in an increasingly digital society, but unmonitored access creates preventable risks. Schools, he suggested, should play an active role in this supervision, requiring parents to seek educational interventions from their children's institutions when concerning patterns emerge.
The minister also raised alarm about the weaponisation of AI in facilitating cyberbullying, particularly within the Tamil-speaking community on platforms like TikTok. He detailed instances where AI-driven content moderation systems have proven inadequate at detecting harassment because they fail to grasp linguistic nuance and cultural context. TikTok's content regulation algorithms, Fahmi explained, lack sufficient training in regional languages and dialects to identify subtle forms of bullying or targeted harassment. This technological failure transforms platforms into zones where vulnerable communities experience coordinated abuse with minimal detection or intervention from automated systems. The issue becomes more acute given that TikTok Live sessions can broadcast for extended periods, giving harassers extended windows to target individuals while moderation systems operate with insufficient linguistic sophistication.
Fahmi's ministry has initiated regulatory pressure on technology platforms through two newly implemented codes: the Child Protection Code (CPC) and Risk Mitigation Code (RMC). These frameworks establish expectations for how platforms must operate to safeguard Malaysian users, particularly minors. The Communications and Multimedia Commission (MCMC) has undertaken international engagement, including discussions in Singapore with major platforms including YouTube, to ensure comprehensive adoption of these standards. The regulatory timeline deliberately provides platforms with several months to restructure their systems, demonstrating government intent to move from advisory warnings toward enforceable obligations backed by significant financial penalties.
The threat of substantial court-imposed fines serves as the enforcement mechanism. Once the CPC and RMC reach full implementation, platforms that fail to maintain adequate safeguarding systems face liability for criminal activities occurring on their services, including fraud, illegal gambling, cyberbullying, and exploitation. This represents a meaningful escalation from previous regulatory approaches that relied on voluntary compliance. Malaysian authorities have signalled that leniency toward non-compliance will evaporate once transition periods expire, placing genuine pressure on global technology companies to invest in better moderation, language understanding, and child protection features.
The Kuala Lumpur Federal Territory Safe Internet PTA Convention 2026, organised by the GoStrong Malaysia Foundation in partnership with MCMC, brought together nearly 500 participants representing parents, educators, school administrators, government agencies, and community organisations. This coalition-building approach reflects recognition that internet safety cannot rest solely on regulatory pressure or parental vigilance. Schools occupy a critical position as institutions that interact daily with young people and can identify concerning behavioural changes. Teachers and parent-teacher associations can serve as early warning systems, detecting when children display signs of unhealthy AI attachment or victimisation through online harassment.
Fahmi's positioning of schools as intervention points reflects a three-layered safety model: parental monitoring at home, institutional oversight within educational settings, and regulatory enforcement at the platform level. This distribution of responsibility acknowledges that no single actor possesses sufficient visibility or authority to protect children across all digital touchpoints. Parents operating without institutional support often lack the technical knowledge to monitor sophisticated AI interactions effectively. Schools lacking regulatory backing cannot enforce standards beyond their campuses. Platforms resisting government pressure can indefinitely delay improvements that protect users. Only through coordinated action across these domains can Malaysia hope to create digital environments where young people can benefit from technological innovation without becoming prey to its darker potentials.
The minister's emphasis on language-specific content moderation reflects a Southeast Asian reality that global technology companies frequently overlook. Platforms designed primarily for English-speaking users and trained on predominantly Anglophone data often struggle with regional languages and dialects. Tamil, Malay, Mandarin, and other languages prevalent in Malaysia and the region carry cultural specificities and contextual meanings that English-trained AI systems cannot reliably navigate. This technological gap has real consequences: harassment in Tamil-language TikTok sessions goes undetected because algorithms trained on English-language harassment patterns cannot recognise analogous abuse conducted in different linguistic registers. Addressing this gap requires platforms to invest significantly in language-specific training data and moderation capabilities, investments that become financially justifiable only through regulatory obligation or market pressure.
For Malaysian parents and educators, Fahmi's warnings suggest an urgent need to develop new digital literacy competencies. Parents must learn not only how to monitor their children's internet activity but also how to discuss AI's limitations intelligently with young people. Educational curricula should incorporate units on AI's capabilities and constraints, helping students develop realistic expectations about technology's role in their lives. Teachers should receive training on recognising signs of unhealthy AI attachment or cyberbullying victimisation. These practical steps complement the regulatory frameworks that will constrain platform behaviour, creating a comprehensive ecosystem oriented toward protecting Malaysian young people in an era where artificial intelligence has become woven into everyday digital experience.
