This is not a marginal or purely local debate. Around the world, artificial intelligence is already entering newsrooms — not only as a risk, but also as a tool for better journalism. It can assist with transcription, translation, data analysis, document review, accessibility and visual storytelling. Even more significant is its role in research, fact-checking, automated monitoring and verification. AI can help journalists process large volumes of information, detect patterns, compare sources, identify manipulation, follow public narratives and expose disinformation more effectively. When used under human editorial control, it can free more time for reporting, investigation, context and editorial judgment. But when regulated through vague, defensive or outdated language, it can also create confusion, fear and mistrust. That is why Article 16 of the new Journalists’ Code in North Macedonia deserves attention beyond the country itself: it is a small example of a much larger global challenge – how to protect journalism from manipulation without stigmatizing the responsible use of technology.

Reading the new “Code of Journalists of Macedonia”, whose drafting, according to the announcement, took two years and was supported by expert analyses and an inclusive process, one provision in particular draws attention: the article concerning the use of artificial intelligence in journalism. The drafting of the new Code was financed by the Government of Switzerland through the Civica Mobilitas programme.

One of the novelties in this new document regulating the work of women and men journalists concerns the use of artificial intelligence. Is the wording of this article in the Code adequate?

According to the English translation made available by the Association of Journalists of Macedonia, Article 16 reads:

Journalistic content created using artificial intelligence must be clearly identified as such. Journalists remain fully responsible for the accuracy of the information and the integrity of the media product.

The use of artificial intelligence for the purposes of creating or disseminating false, manipulative, or substantially altered content that may mislead the public is unacceptable. The journalistic product must not consist entirely of content generated by artificial intelligence.

This is a necessary adjustment to the new digital era, in which artificial intelligence has become an integral part of human life as a whole. The good intentions are obvious. However, the wording of this article in the Code is too general, and the tone toward artificial intelligence is defensive — as if artificial intelligence were a threat, and not also a tool in the work of a newsroom.

Truth, verification, independence, the public interest and responsibility are understood as fundamental. Technology must not replace professional judgment.

“Journalists remain fully responsible for the accuracy of the information and the integrity of the media product.”

All of this is good. After all, this is precisely what artificial intelligence platforms and models themselves say — from ChatGPT to Gemini and others: their outputs may be inaccurate, incomplete or wrongly contextualized, and therefore must not replace human verification, professional judgment and editorial responsibility. Paradoxically, while the Code takes a defensive stance toward artificial intelligence, it essentially adopts the very message that AI tools themselves constantly repeat: outputs must be checked, contextualized and placed under human and editorial responsibility.

The Code also rightly prohibits the use of artificial intelligence for false, manipulative or materially altered content that misleads the public. This is also in line with the broader European and international regulatory framework, such as the Council of Europe Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law (Vilnius, 2024), as well as the EU Artificial Intelligence Act (AI Act). These frameworks emphasize the responsible use of artificial intelligence across all stages of journalistic work.

In the context of European regulation, the provision that “journalistic content created using artificial intelligence must be clearly identified as such” is too unclear.

Does this include transcription? Translation? Grammar correction? Summarizing a long public document? Consulting AI when drafting headline options? Data analysis? If so, then almost everything could require a label, creating meaningless “AI-assisted” labels virtually everywhere. That would not increase trust in the media; it would only produce fatigue and make media content less clear and less navigable.

A standard that could be both ethical and transparent would require such labeling when artificial intelligence materially contributes to the published content, especially through text, image, audio, video, data visualization or synthetic reconstruction that the audience could reasonably interpret as AI participation in the presentation of journalistic evidence.

The final sentence of this article in the Code is also problematic.

According to the English translation made available by the Association of Journalists of Macedonia, the sentence reads:

“The journalistic product must not consist entirely of content generated by artificial intelligence.”

The intention behind this provision is entirely acceptable, but the Association of Journalists of Macedonia has an obligation to be more precise. “Content generated entirely by artificial intelligence” is certainly something that cannot and should not be published as journalism. However, an illustration generated by artificial intelligence, clearly identified as an illustration, may be acceptable. An automated table of election results or a weather forecast report may also be acceptable, provided that it is based on verified data and editorial oversight.

The question is not only whether artificial intelligence generated the content, but whether there was human editorial control, verification, sources, responsibility and transparency.

“Generated by artificial intelligence” is not the same as “assisted by artificial intelligence.” A ban on a journalistic product that “consist[s] entirely of content generated by artificial intelligence” may sound clear at first glance, but in fact it avoids the more difficult questions: What counts as “entirely”? What about automated data journalism? What about AI-assisted visualizations? What about a human-edited report generated from verified official datasets? What about clearly labeled illustrations? These questions are not answered by the rigid wording in the Code.

Many serious questions remain unanswered, although European institutions, as well as researchers working on the ethical application of artificial intelligence, already offer at least partial answers. Two examples follow, the second of which is particularly important.

More serious protection against synthetic content is also needed: AI-generated images, voices or videos must never be presented as authentic documentation.

Furthermore, protection of privacy, copyright and sources is also needed. The article in question does not cover this, although it is exceptionally important. Journalists should not upload confidential material, personal data, information about whistleblowers or unpublished investigative material into unsafe artificial intelligence tools. This is far more important than merely listing several points that AI platforms themselves constantly emphasize.

This part of the Code, Article 16, must be seriously revised, as should a number of other provisions in this important document. As written, Article 16 represents only a rather unsuccessful attempt to protect the public from the abuse of artificial intelligence. At the same time, it creates a real risk of stigmatizing the legitimate use of artificial intelligence in newsrooms. It will cause confusion, while doing very little – if anything – to strengthen public trust in the media. And all of this has been elevated to the level of a Code.

In short, this provision in the new Journalists’ Code recognizes the risk, but not the reality. It rightly insists on responsibility, accuracy and transparency, but fails to distinguish between journalism generated by artificial intelligence, AI-assisted newsroom work, synthetic manipulation, translation, transcription, research support, headline suggestions, data processing, visual illustration, summarization or accessibility tools. By placing all of these under the same vague umbrella, it creates confusion instead of standards. And mistrust can only deepen.

What is needed is a genuine understanding of the global digital reality, especially in the context of the media industry. What I am reading now is merely an attempt to regulate a new technological reality with old – backward – reflexes that treat artificial intelligence almost as a contaminant in journalism, instead of treating it as a tool whose ethical application depends on how, where, why and under whose responsibility it is used.

YOUR SUPPORT TO CIVIL MATTERS.

Support independent journalism, democratic resilience, and the defense of freedom of expression. Help CIVIL counter disinformation, hybrid threats, and authoritarian influence.

Contribute NOW