Malaysia’s Media Council puts human oversight and AI labels at heart of draft ethics code

Malaysia’s Media Council is developing guidelines for responsible AI communication that would keep editorial responsibility with journalists and require clear labels for relevant synthetic content. The proposed rules place human ethics, judgment and oversight as the final authority, even as AI tools become more capable. An opinion article published by The Vibes describes the initiative as a priority shared with the Ministry of Communications within the council’s self-regulation framework. Its central principle is that AI should assist rather than replace journalists, who remain responsible for the accuracy and fairness of their work regardless of the technology used. The guidelines identify human-led ethics, transparency and labelling, alignment with Malaysia’s National AI Roadmap, and harmonisation with international standards as their main pillars. They would require “AI-generated” or “AI-enhanced” tags for applicable material, responding to concerns about deepfakes and platform rules, and cite the Venice Pledge as a model for ethical, human-centric AI communication. The framework would operate through the council’s Code of Conduct and Complaints Committee, which handles public complaints involving media ethics, AI transparency and journalistic standards. It is a broad, outward-facing and co-regulatory approach: one intended to curb local misinformation, protect national discourse and give the public a route to complain when outlets fail to disclose synthetic media or breach ethical standards. The council was enabled by the Malaysian Media Council Act 2025, which took effect on June 14, 2025. Twelve founding members were appointed, and the body’s statutory functions include setting a journalists’ code of conduct, promoting media freedom, regulating professional ethics, engaging the public and government, and recommending media-law reforms. The AI proposal sits alongside Malaysia’s wider national policy. The government’s National Guidelines on AI Governance and Ethics call for transparency, accountability, fairness, privacy, security and human rights, while setting out seven principles including reliability, safety and control, inclusiveness and human benefit. Malaysia’s National AI Roadmap covered 2021 to 2025 and included governance, ethics, cybersecurity, standards and responsible-AI impact assessment. The article draws a distinction between the council’s public-facing standards and detailed rules that individual newsrooms may need. Its proposed newsroom framework calls for human accountability, verification of AI-produced claims against primary sources, disclosure of significant AI alteration, privacy and source protection, and audits for algorithmic bias. Such operational rules would also address technical practices beyond the council’s broad framework. They would prohibit sensitive inputs, unreleased drafts and off-the-record information from being entered into inadequately encrypted public cloud systems; require sensitive transcription, translation and analysis to occur on locally encrypted offline hardware; and restrict AI imagery to clearly labelled illustrative or educational use rather than fabricated breaking-news photographs or video. The article points to Complaint F2-013 to illustrate the limits of external recourse. A journalist was publicly criticised over an AI-generated image, and received a show-cause letter from their employer even after corrective action. The council secretariat found it lacked jurisdiction because the respondent was neither a media organisation nor a council member and the complaint concerned a social-media post, though it continued limited engagement with the employer about the letter. The article argues that this episode exposed a gap between accountability standards and protection for individual journalists. It also raises concerns that the AI code is being developed largely through top-down focus groups with limited scope for wider stakeholder submissions, while identifying plagiarism risks from AI systems drawing on internet material without adequate attribution or originality safeguards as insufficiently addressed.