Policy Brief #98 - Regulating AI in the media: supporting newsrooms, translating rules, and closing the gap with editorial practice

3 November 2026

Aina Errando

Artificial intelligence (AI) is now part of everyday news work, from transcription and translation to verification, recommendation, and content generation. At the same time, Europe has built one of the most extensive AI governance frameworks in the world, combining AI-specific rules with platform, media, data protection, and copyright law. Most of these rules are horizontal: They apply across sectors and say little about what makes journalism distinct, from editorial responsibility to its democratic role. Between April and October 2026, we interviewed 18 regulators, legal experts, and news professionals at national regulatory authorities and ministries in Belgium, France, and Spain, at European and international organisations, and at public service and commercial media organisations. This brief examines how AI regulation reaches news organisations in practice, how far it shapes the way they govern AI, and where policy should act next. 

1.    Why AI governance matters for the media

In most European countries, the media set their own ethical standards, and governments stay out of editorial choices. That model is now under pressure from three directions. The systems newsrooms use are largely developed and controlled by technology companies that journalistic codes cannot reach. Their risks, from synthetic media to the unauthorised use of news content for training, extend beyond individual stories to the information ecosystem as a whole. Meanwhile, international bodies and legislators have adopted a growing set of rules and guidelines. For instance, the Council of Europe has issued guidelines on the responsible use of AI in journalism¹, and the AI Act imposes obligations on news organisations as deployers of AI systems². Newsrooms, in turn, have responded with their own guidelines, which increasingly resemble one another³. Little is known, however, about how European rules actually reach newsrooms, whether they change how news organisations govern AI, and which mix of self-regulation, co-regulation, and binding regulation those involved consider appropriate.

2.    Where AI law meets the media

Across levels of governance, interviewees described AI regulation as horizontal: designed to manage the risks of a technology rather than to address the institutions that use it. Supranational experts were most critical, arguing that current frameworks focus on markets, competition, and privacy while overlooking journalism’s role in the information ecosystem. National regulators, too, acknowledged that the rules were not written with the media in mind.

The few points where AI law meets journalism concern general issues rather than news as such. Regarding transparency, the AI Act asks for AI-generated text on matters of public interest to be labelled, except where a human has reviewed it and someone carries editorial responsibility for it. When it comes to copyright, the most pressing concerns in news media organisations are the use of their content to train AI models and the lack of remuneration, issues that arise outside the newsroom. Interviewees also stressed that regulation moves too slowly, so that guidance risks being outdated by the time it is published.

3.    Who puts AI rules into practice in news organisations

Legal teams as translators. Few journalists deal with AI law first-hand. Inside news organisations, it is the privacy and legal specialists, sometimes helped by external lawyers, who turn the rules into in-house policies. Several interviewees doubted that most journalists know the rules at all. Regulators, in turn, acknowledged their own uncertainty about how general provisions apply to the media, and described their role as translating a complex framework for the sector.

Guidelines shaped by peers and ethics codes. The AI Act was seldom the reason newsrooms wrote AI guidelines in the first place. Charters were modelled on those of other media, such as the BBC, the New York Times, or The Guardian, co-written with journalists’ associations, or anchored in existing deontological and ethics codes.

Gaps between guidelines and daily work. Guidelines on paper do not guarantee practice. Interviewees described the use of unauthorised AI tools (“shadow AI”), charters that age faster than they are revised, and limited understanding of how AI tools work. Being transparent about AI use is not the same as being able to explain it to audiences.

4.    When the law mirrors newsroom practice

Where binding rules do touch journalism, they largely codify norms that newsrooms already hold. The clearest case is editorial responsibility: In every news organisation we studied, a journalist remained accountable for anything produced with AI, often before any regulation asked for it, and the interviewed outlets experience the law as confirming their practice rather than constraining it.

The mirroring protects editorial autonomy, but it has costs. It lets news organisations define what compliance means: Disclosure thresholds were set by newsrooms themselves, and some AI uses, from spelling correction to live subtitling, were deliberately left unlabelled. Moreover, outlets without a style book or an editorial statute have little to fall back on, and translating horizontal rules requires expertise that mainly large organisations can afford.

5.    Which kind of regulation for which issue

Stakeholders did not favour one regulatory approach across the board but matched it to what is being governed. Journalists preferred self-regulation for how they use AI in their own work, where editorial independence is at stake. Co-regulation was associated with shared standards, such as AI and content provenance labels, yet no forum currently exists in which media, regulators, and technology firms could agree on them. Binding regulation was considered indispensable for synthetic media and disinformation, and for the companies that build AI systems. What made regulation acceptable was less how much the state intervenes than whose conduct is regulated.

6.    Conclusion and upcoming challenges

For now, European AI law mostly mirrors the professional rules that newsrooms already apply. While this may protect editorial autonomy, it leaves the most pressing issues unaddressed: journalism's dependence on a few AI tool providers, the use of news content without remuneration, and the lack of shared transparency standards. It also favours large organisations that can afford data protection officers and legal advice, while smaller outlets without editorial codes are left to navigate the rules alone.

The task for policy is therefore not to add rules for newsrooms, but to clarify how existing editorial rules apply to AI, to build the missing co-regulatory infrastructure, and to ensure that responsible AI use is within reach of smaller outlets as well as large ones.

7.    Recommendations

DOWNLOAD THE POLICY BRIEF

About ALGEPI

This policy brief summarises findings from the ongoing project ALGEPI. ALGEPI is funded by the Research Foundation – Flanders (FWO) under grant number G098223N.