1. From a Facebook post to a news story
Can a Facebook post be protected by copyright? If so, how much of it may the press reproduce when reporting a current event?
These questions brought a dispute before the Court of Justice in Gândul Media, C-598/24. A Romanian schoolteacher published a 22-line Facebook post entitled «Short guide for parents at the beginning of the school year», explaining why she did not wish to receive gifts from pupils’ parents. Six days later, an online newspaper reproduced the post in full without her prior consent. Her name and a hyperlink to the original post were added subsequently.
On 3 September 2026, the Court held that a short text posted on a social network can constitute a copyright work where it expresses the author’s own intellectual creation. It also clarified the discretion available to Member States when implementing the exception for reporting current events under Article 5(3)(c) of Directive 2001/29.
Yet the judgment is interesting beyond Facebook posts and online newspapers. Its emphasis on informational purpose, the amount reproduced and whether reproduction replaces consultation of the original touches upon a persistent problem in EU copyright law: when does an intermediary facilitate access to information, and when does it become a substitute for the source?
That question has already surfaced in disputes concerning search engines, news snippets and the press publishers’ right. It takes on renewed importance as search increasingly moves towards generative AI systems capable not merely of locating sources, but of supplying the answer themselves.
2.Another chapter in the Court’s notion of a work
The Court’s approach to the notion of a copyright work is now well established. From Infopaq, C-5/08, through Levola Hengelo, C-310/17, Cofemel, C-683/17 and Brompton Bicycle, C-833/18, to the more recent joined cases Mio/Konektra, C-580/23 and C-795/23, the Court has consistently treated the notion of a work as an autonomous concept of EU law.
Protection requires two cumulative elements: (i) original subject matter constituting the author’s own intellectual creation; (ii) an expression of that intellectual creation that is identifiable with sufficient precision and objectivity. Originality depends upon the author being able to make free and creative choices through which their personality is reflected in the resulting work.
Gândul Media does not significantly alter this framework. Rather, it applies it to another form of everyday expression. The length of the text, its publication on a social network and the absence of a predetermined literary genre are not decisive. In written material, creativity may be reflected through the selection, arrangement and combination of words. The final determination remains with the national court, but the facts before the CJEU suggested that the teacher had sufficient creative freedom when drafting her post.
The judgment therefore confirms that the informal nature of social media expression does not, in itself, exclude it from copyright protection. At the same time, an important boundary remains: copyright protects expression, not the underlying opinions, facts or ideas themselves.
3.Reporting current events: how much is too much?
The central issue in Gândul Media concerns Article 5(3)(c). The provision permits uses of protected works in connection with reporting current events, to the extent justified by the informatory purpose and subject to identification of the source, including the author’s name, unless this proves impossible. In Spiegel Online, C-516/17, the Court had already interpreted the exception in light of freedom of expression and information under Article 11 of the Charter.
The Court also addressed what may count as reporting a «current event». It held that reporting involves providing information about an event of current informational interest, without requiring detailed analysis or an invitation to public debate. A teacher’s comments on school practices at the beginning of the school year could therefore fall within that concept, although the final determination remains with the national court.
Romanian law imposed two additional conditions: (i) reproduction was limited to short extracts, and (ii) could not generate any direct or indirect commercial or economic advantage.
The Court held that Article 5(3)(c) does not preclude Member States from restricting the exception to short extracts, provided that such a restriction is proportionate, preserves the effectiveness of the exception and permits observance of its informatory purpose. The provision itself establishes no quantitative threshold, but freedom of information and freedom of the press do not invariably require reproduction of an entire work.
The Court connected this assessment with the three-step test in Article 5(5) of Directive 2001/29. Particularly noteworthy was its concern that full reproduction may replace the original communication of a work, making it unnecessary for the public to consult the original. This was not a categorical prohibition on full reproduction. Very short works may make extraction impracticable. Nevertheless, the relationship between the secondary communication and the original becomes relevant to the copyright balance.
By contrast, the Court rejected the blanket prohibition on commercial or economic benefit. Professional media ordinarily pursue economic activities while also informing the public. Making non-commerciality a general condition would undermine the effectiveness of the exception and disturb the balance between IP under Article 17 of the Charter and freedom of expression and information under Article 11.
Commerciality is therefore not determinative. The assessment turns instead on the purpose of the use, its proportionality and the extent of the reproduction.
4.A familiar problem: search, snippets and publishers
The concern in Gândul Media about whether a secondary use replaces consultation of the original is not new. Similar tensions have long characterised the relationship between search engines, news aggregators and press publishers.
In VG Media v Google, C-299/17, the dispute concerned Google’s display of extracts from press publications in Google Search and Google News under an earlier German press publishers’ right. Although the CJEU ultimately resolved the case on a notification issue, the broader publisher-platform conflict formed part of the background to Article 15 of the CDSM Directive.
Article 15 grants press publishers reproduction and making-available rights for online uses of their publications by information society service providers, while excluding hyperlinks, individual words and «very short extracts». What constitutes a «very short extract», however, remains uncertain. Recital 58 cautions against interpreting the exclusion so broadly that the right becomes ineffective, without establishing a quantitative threshold.
The literature has therefore treated the issue as more than one of word count: the question is when an extract facilitates access to information without undermining the publication itself (Rendas, 2022). National implementation has also produced divergences despite Article 15’s harmonising objective (Furgał, 2023, Rosati, 2026).
Article 15 concerns different rights and right holders from Gândul Media. Yet both expose a related practical problem: when does an intermediary cease to facilitate access to a source and begin to replace it?
5.From snippets to generated answers
Generative AI gives that problem a new dimension. Traditional search engines principally direct users towards external sources through titles, hyperlinks and snippets. Generative search can instead synthesise those sources into a self-contained answer, potentially reducing the need to visit them.
This shift has now received regulatory recognition outside copyright. On 31 August 2026, the European Commission designated ChatGPT as a Very Large Online Search Engine under the Digital Services Act (European Commission, 2026). The designation does not determine any copyright question, but it places ChatGPT’s relevant search functionality within the DSA’s online-search-engine framework. As Bassini and Palumbo argue, generative AI sits uneasily within the DSA’s traditional taxonomy of intermediary services. While some applications functionally resemble online search engines or platforms, their capacity to generate outputs themselves complicates their classification as mere intermediaries and raises broader questions about whether existing legal categories remain adequate (Bassini and Palumbo, 2026).
For copyright, the significance lies in the changing form of intermediation. Earlier disputes focused largely on how much of a source could be displayed while still directing users towards the original. Generative systems raise a different problem: the intermediary may synthesise information drawn from multiple sources and present the resulting answer as a self-contained informational product.
6RAG: from retrieval to replacement?
This issue is particularly visible with retrieval-augmented generation, or RAG. RAG enables a generative system to retrieve external material relevant to a query and condition its response on that material rather than relying exclusively on information encoded within the model (Lewis et al., 2020). For current events, those sources may include newspaper articles or social media posts published only minutes or hours earlier.
This differs from the familiar debate about copyrighted works used for AI training. As the European Copyright Society (ECS) has emphasised in its Opinion on Like Company v Google, RAG is generally a deployment-stage technique, in which external content is dynamically retrieved at inference time rather than incorporated into the model’s learning process (ECS, 2026). The copyright inquiry therefore shifts towards retrieval and output: a protected source is selected because it contains information responsive to a particular query and is then used in constructing an answer.
Existing scholarship has already identified difficulties in applying the EU press publishers’ framework to generative AI uses of press publications (Kowala, 2024). More recent work has questioned whether the publisher-AI conflict should be approached primarily through stronger proprietary rights or through broader concerns of fairness and the functioning of information markets (Kraetzig, 2025).
Gândul Media does not establish that AI systems using RAG are «reporting current events», nor that Article 5(3)(c) provides an exception for generative search. Those questions remain open. Its reasoning nevertheless offers a useful lens.
This matters increasingly as generative AI becomes embedded within everyday information services. Users seeking information about a breaking news story may now encounter an AI-generated summary before reaching the underlying sources themselves. Google Search, for example, increasingly integrates generative responses through AI Overviews and AI Mode, which can synthesise information from multiple webpages and present it directly within the search experience. Google reported in 2026 that AI Overviews had surpassed 2.5 billion monthly active users (Google, 2026), illustrating the scale at which generated answers are becoming embedded in ordinary information seeking.
The copyright relationship may therefore differ depending on how that answer is constructed. A response that briefly summarises a development and directs users towards the underlying reporting has a different relationship with the protected material from one that reproduces or synthesises its substance so comprehensively that consulting the original sources becomes unnecessary. One relevant question is therefore whether the generated answer facilitates access to protected sources or functionally substitutes for them.
This is where Gândul Media connects most clearly with the earlier debate around snippets and press publishers. Article 5(3)(c) asks whether reproduction is justified by an informatory purpose; Article 5(5) requires consideration of normal exploitation; and Article 15 draws a boundary around «very short extracts». These provisions address different rights and legal questions, but they reveal a recurring concern about the relationship between secondary uses and the sources on which they depend.
As generative search increasingly mediates access to news and other current information, Gândul Media offers a useful reminder that the copyright inquiry may turn not only on what is reproduced, but also on the relationship between the intermediary’s output and the original source.
