Malaysian parents need to exercise greater vigilance when their children begin treating artificial intelligence systems as companions or romantic partners, Communications Minister Datuk Seri Fahmi Fadzil warned on Thursday. Speaking at the Kuala Lumpur Federal Territory Safe Internet PTA Convention 2026, Fahmi highlighted the psychological risks posed by conversational AI technology that mimics human interaction without possessing human qualities or moral judgment. His intervention reflects growing concern within Malaysia's regulatory and parental communities about the normalization of AI relationships among young people, particularly as these systems become increasingly sophisticated at simulating emotional connection.
While acknowledging that AI tools offer legitimate educational and informational benefits, Fahmi drew a critical distinction between using technology for practical purposes and developing emotional dependency on it. The crux of his warning centred on the deceptive nature of advanced chatbots, which are engineered to respond in conversational patterns that closely resemble human dialogue. However, these systems operate on patterns extracted from human-generated data and reflect neither genuine understanding nor authentic concern for the user. The minister emphasized that parents must recognize this fundamental difference and help their children internalize it, preventing situations where young people seek emotional validation or companionship from algorithms rather than from trusted adults or peers.
A particularly concerning dimension of AI interaction that Fahmi identified involves the potential for these systems to provide false or outdated information presented with apparent confidence. Unlike human sources who can acknowledge uncertainty or admit knowledge gaps, AI models sometimes generate plausible-sounding but entirely fabricated information—a phenomenon technologists call "hallucination." This creates a double risk for young users: they may internalize incorrect facts while simultaneously trusting the source because it communicates in a friendly, seemingly knowledgeable manner. The minister cautioned that the data powering these systems may be months or years old, meaning even when functioning correctly, they cannot reliably address current events or rapidly evolving situations.
Rather than imposing blanket prohibitions on internet use—an approach Fahmi acknowledged would be counterproductive in an increasingly digital world—he advocated for informed parental oversight and engagement. Parents should maintain awareness of their children's online activities without becoming intrusive, while schools can serve as partners in digital literacy education. This balanced approach recognizes that complete internet deprivation is neither realistic nor beneficial, as digital skills have become essential for academic and professional success. Instead, the emphasis shifts to ensuring that children develop critical thinking about their online interactions and understand the distinction between entertainment, information, and genuine human connection.
Fahmi's broader concerns extend beyond AI relationships to encompass the role of algorithmic content moderation in protecting vulnerable communities online. He raised alarms about cyberbullying within Tamil-speaking communities on TikTok, where coordinated harassment campaigns have escalated despite platform policies against such behavior. This issue highlights a critical gap in content moderation: artificial systems, particularly those relying on machine learning, often fail to comprehend linguistic nuance, cultural context, and coded language used in harassment campaigns. Tamil speakers, like other minority linguistic communities, may employ expressions or references that algorithms cannot adequately analyze, leaving bullying tactics undetected while legitimate speech might be incorrectly flagged.
The minister disclosed that he had directly engaged TikTok regarding these failures, pressing the platform to recognize that language moderation cannot be entirely automated when cultural sensitivity and contextual understanding are required. Human moderators with genuine fluency in Tamil and familiarity with Tamil cultural references are essential for identifying harassment that would sail past algorithmic filters. This represents a significant challenge for global platforms operating across dozens of languages and cultural contexts, as hiring and training sufficient numbers of qualified moderators for every linguistic community remains economically challenging and organizationally complex.
Fahmi's comments also addressed Malaysia's regulatory framework for digital platforms, specifically the implementation of the Child Protection Code (CPC) and Risk Mitigation Code (RMC). These instruments represent the government's effort to establish binding standards for how social media and digital platforms operate within Malaysian jurisdiction, covering issues from fraud and online gambling to cyberbullying and child safety. The Malaysian Communications and Multimedia Commission (MCMC) has been conducting discussions with major platforms including YouTube in Singapore, establishing timelines for compliance and ensuring that operators understand their obligations. The government has provided platforms several months to reconfigure their systems, but the timeline is not indefinite.
The stakes for non-compliance are substantial. Once the CPC and RMC are fully implemented, platforms that fail to meet their requirements face potential court proceedings and penalties of sufficient magnitude to motivate genuine operational changes. This enforcement mechanism is designed to move beyond voluntary corporate compliance toward legal accountability, mirroring approaches adopted in Europe and Australia. For Malaysian users and parents, this means that digital platforms will eventually operate under clearer, locally-enforced standards rather than solely adhering to global policies that may not address Malaysia-specific concerns.
The convention itself represented a significant gathering of stakeholders from across Malaysia's education and digital safety ecosystem. Nearly 500 participants included parents, teachers, school administrators, Parent-Teacher Association leaders, government agency representatives, and civil society organizations. This multi-stakeholder approach reflects an emerging consensus that digital child safety cannot be addressed by any single sector alone. Schools, families, government regulators, and technology companies must coordinate their efforts, share information about emerging risks, and develop complementary strategies. The participation of such a diverse group indicates that Malaysian institutions are moving beyond awareness-raising toward building actual implementation capacity.
For Malaysian families specifically, Fahmi's warnings arrive at a moment when AI applications are becoming more widely accessible and affordable. ChatGPT, along with numerous competing systems, are freely available to anyone with internet access, and younger Malaysians are adopting these tools at increasing rates. The risk that he identifies—that children may form unhealthy attachments to AI systems, mistake automated responses for genuine friendship, or internalize false information presented persuasively—is not hypothetical. Case studies from other countries document instances where young people have become emotionally dependent on chatbots or, in extreme cases, engaged in interactions with AI systems that reinforced harmful behaviors or ideation.
The broader developmental concern is how early exposure to artificial intimacy might shape children's expectations about relationships and trust. If young Malaysians become accustomed to interacting with systems that are designed to be non-judgmental, endlessly patient, and constantly available, they may develop unrealistic expectations about human relationships, which inherently involve conflict, disappointment, and the need for genuine effort and compromise. Conversely, they may become disillusioned with human friendships that do not match the frictionless dynamics of AI interaction, potentially increasing isolation and mental health challenges.
Fahmi's intervention also reflects Malaysia's broader positioning on technology governance, distinct from both unregulated permissiveness and authoritarian control. The government is attempting to establish baseline standards that protect vulnerable populations—particularly children and minorities—while preserving the benefits of digital innovation and internet freedom. This middle path requires ongoing dialogue between regulators, industry, educators, and families. The government's emphasis on parental involvement and school-based intervention, combined with platform accountability through codes of conduct and enforcement mechanisms, suggests a recognition that digital child safety is fundamentally about sharing responsibility rather than concentrating it in any single authority.
Looking forward, Malaysia's approach to AI and digital safety will likely influence how other Southeast Asian nations think about similar challenges. As a middle-income country with substantial digital adoption and a diverse, multilingual population, Malaysia's regulatory experiments and policy frameworks may offer valuable models for regional neighbors. Fahmi's warnings about AI friendships, combined with the government's enforcement of content standards and platform accountability, represent an attempt to establish that technology companies operating in Malaysia must meet local expectations about user protection, particularly for minors.
