A COMPREHENSIVE code of Artificial Intelligence ethics for journalists aims to adapt traditional editorial values in an algorithmic era.
The Society of Professional Journalists in the United States has long held that while technology evolves, one core tenet should remain fixed.
Journalists must be ultimately responsible for the accuracy and fairness of their work, regardless of the tools used.
In Malaysia, that principle is now being formalised under the Malaysian Media Council.
The MMM, established under the Malaysian Media Council Act 2025, which designates it as an independent statutory body, has a mission of overseeing print, broadcast, and digital platforms. Its stated founding principle is that a free and responsible press forms the cornerstone of democracy, and it claims to operate independently of government and commercial interests.
However, the MMM’s sources of income are completely opaque.
The MMM’s formal mandate includes ethical standard-setting, complaints handling, dispute mediation, capacity development, and international representation of the Malaysian media.
However, some pointed concerns arise from its operational emphasis.
Critics note that the council appears more focused on investigating and adjudicating complaints against journalists than on shielding them from external pressures.
Journalistic NGOs whose missions centre on protecting reporters have aligned with the MMM, even as the body has been slow to address abuse, Strategic Lawsuits Against Public Participation (SLAPP) actions designed to silence critical coverage, or the steady decline in the number of skilled journalists, a trend many attribute to growing rigidity and control within the industry.
One has to worry that the MMM risks becoming both investigator and judge in cases brought against journalists.
A recent illustration appears in Complaint F2-013 (listed on the MMM website).
In this case, a journalist faced public criticism over an AI-generated image in an article.
After corrective action was taken, the journalist still received a show-cause letter from their employer.
Attempts to have the original social-media post removed failed.
The MMM Secretariat ultimately decided it lacked jurisdiction because the respondent was neither a media organisation nor a member, and the matter involved a social-media posting rather than a formal media outlet.
While noting the reputational damage for the record, the Secretariat continued limited engagement with the employer’s person-in-charge regarding the show-cause letter.
The episode with the MMM exceeding legislative jurisdiction underscores how AI-related controversies can leave individual journalists exposed while institutional boundaries limit the council’s protective reach.
Working in tandem with the Ministry of Communications, the MMM has prioritised integrating AI ethics into its self-regulation framework through Guidelines for Responsible AI Communication.
The core directive stresses that human ethics, judgment, and oversight must remain the final authority even as machine capabilities advance.
Key pillars include human-led ethics (AI as assistant, not replacement), transparency and content labelling (explicit “AI-generated” or “AI-enhanced” tags, driven partly by concerns over deepfakes and stated platform rules), alignment with Malaysia’s National AI Roadmap, and harmonisation with international standards such as the Venice Pledge for ethical, human-centric AI communication.
These guidelines operate under the 2025 Act via the council’s Code of Conduct and Complaints Committee, which mediates public complaints on media ethics, AI transparency, and journalistic standards.
The framework is deliberately co-regulatory and macro-level, where it seeks to prevent localised misinformation, protect national discourse, and give the public a standardised channel for complaints when outlets fail to label synthetic media or breach ethical boundaries. Infrastructure and privacy details are largely left to individual media organisations under Malaysian digital-data laws.
In contrast, more granular internal newsroom standards focus on day-to-day technical realities.
They warn against automation bias (passively accepting algorithmic outputs under deadline pressure), require correction synchronicity (updating internal AI knowledge bases so retractions propagate), ban cloud-hosted consumer large language models for sensitive inputs, mandate local offline processing on encrypted hardware, and prohibit the creation of fake breaking-news photography or video, restricting AI imagery to clearly labelled illustrative or educational uses.
Accountability in such codes is inward-facing, relying on routine interdisciplinary audits rather than external public recourse.
These differences matter. The MMM approach is broader and outward-facing, with potentially tighter operational codes in an already financially strapped industry, that may become more restrictive on infrastructure and synthetic content.
Both share the goal of human oversight, yet their scopes diverge sharply.
Of particular concern is the process by which the MMM’s AI code of conduct is being developed.
A top-down focus-group method appears to predominate, with limited room for open brainstorming or submissions with code enhancements from a wide range of stakeholders.
An effective, actionable framework for modern newsrooms, now being used around the world, rests on five foundational principles that many practitioners already recognise as essential and are practising:
1. The Human-in-the-Loop Imperative (Accountability)
AI extends human capabilities; it does not replace them. Authorship, bylines, and legal liability remain non-delegable. Journalists must guard against automation bias and subject every integrated AI system to continuous auditing by teams that combine technical and ethical expertise.
2. Absolute Accuracy and Verification
Generative tools produce hallucinations. Every claim, timeline, quote, or data point must be cross-checked against primary sources.
Fake breaking-news photography or video is prohibited. Protocols must ensure that corrections and retractions update internal AI knowledge bases so outdated material is not recirculated.
3. Radical Transparency
Audiences deserve clear labels on any text, audio, or visual content significantly generated or altered by AI.
When machine learning is used for investigative data parsing, the logic, inputs, and constraints of the algorithm should be disclosed.
4. Privacy and Source Protection
Sensitive information, unreleased drafts, or off-the-record material must never be fed into public cloud-hosted systems that lack end-to-end encryption.
Transcription, translation, and analysis of sensitive material should occur only on offline, locally encrypted hardware.
5. Algorithmic Fairness and Bias Control
Training data often embeds systemic biases. Tools used for ideation or demographic analysis require active auditing.
Algorithmic suggestions must not homogenise coverage or undermine diversity of sources and community representation.
One further issue has received surprisingly little attention.
That is plagiarism risks arising when AI systems draw on existing internet sources without clear attribution or originality safeguards.
This gap remains largely unaddressed in current discussions.
The MMM’s effort to embed AI ethics into Malaysian media self-regulation is timely. Yet the combination of funding opacity, a complaints-heavy posture that offers limited protection against external pressure, restricted stakeholder input, and unresolved questions around individual journalist accountability leaves a residual sense of unease.
Whether the emerging guidelines will ultimately strengthen or constrain independent journalism depends less on the principles themselves than on how openly, inclusively, and protectively they are implemented. – October 16, 2026