Everyone knows OpenAI; the company behind ChatGPT has become one of the most recognisable names in the technology industry. Yet on 15 July 2026, the General Court of the European Union concluded in OpenAI v EUIPO (Case T-555/25) that the word “OPENAI” simply isn’t distinctive enough to function as an EU trade mark for artificial intelligence-related goods and services. The Court dismissed OpenAI’s appeal and upheld the European Union Intellectual Property Office’s refusal to register the mark under Article 7(1)(b) and Article 7(1)(c) of the EU Trade Mark Regulation.
That outcome might sound surprising. How can a brand recognised by hundreds of millions of people not qualify for protection? But the surprise dissolves once you confront a deeper question: what is trade mark law actually for?
At its core, a trade mark is a legal mechanism that helps consumers identify the commercial origin of goods and services (OpenAI v EUIPO, para. 30). When a sign fails to perform that origin-identifying function — because the relevant public perceives it as describing what the product is rather than who makes it — registration must be refused. The OpenAI judgment therefore exposes the tension between rewarding brand investment and preserving the linguistic commons.
Trade Marks Protect Commercial Origin, Not Technology
To understand why the General Court reached its conclusion, one needs to start with the fundamental architecture of EU trade mark law. Article 7(1)(c) of Regulation 2017/1001 bars registration of signs that consist exclusively of indications which may serve, in trade, to designate characteristics of the goods or services concerned (see para. 32). Article 7(1)(b) separately bars signs that are devoid of distinctive character (see para. 23). These provisions clarify that descriptive terms must remain freely available for all market participants to use.
Stay ahead of the geopolitical week.
MD Briefing delivers expert analysis across five global fronts — the Indo-Pacific, energy, geoeconomics, European security, and the Middle East — every Monday morning. Free.
The rationale for both provisions is the same, but the analytical implications differ. Article 7(1)(c) serves a competition function; it keeps descriptive vocabulary in the public domain so that rival undertakings can describe their own products honestly. On the other hanf, article 7(1)(b) serves a communication function; it prevents the register from being cluttered with signs that consumers simply cannot use as origin identifiers.
The distinction matters because a sign may fail one test without necessarily failing the other. In practice, though, as the OpenAI case illustrates, a finding of descriptiveness under Article 7(1)(c) almost invariably entails a lack of distinctiveness under Article 7(1)(b). The critical question is whether that near-automatic overlap obscures legitimate cases where a descriptive combination has begun to function as an origin identifier despite its semantic transparency. The OpenAI case forces that question into the open.
Why “OPENAI” Became Descriptive: A Journey Through the Case Law
The OpenAI judgment sits at the end of a long line of authority on the registrability of compound signs formed from descriptive components. That line of authority reveals a progressive tightening of the standard, from a comparatively generous early approach to a stricter modern orthodoxy, and a corresponding shift in the Court’s underlying theory of what “descriptive” actually means.
Here’s the basic idea. Trade marks exist to tell consumers who made a product, not what the product is. A company can own a distinctive brand name, but it cannot monopolise ordinary language that competitors may legitimately need to describe their own goods or services. That’s why terms such as “Creamy” for yoghurt or “Fast Delivery” for courier services would generally never qualify as trade marks, they describe a characteristic of the product rather than its commercial origin.
The challenge gets more complex when a trade mark combines two descriptive words into a single expression. At what point does a combination become original enough to identify a business rather than simply describe a product?
The Court of Justice first tackled this question in the case BABY-DRY case in 2001. Procter & Gamble attempted to register “BABY-DRY” for nappies. Although both “baby” and “dry” are ordinary English words, the Court held that putting them together in this particular way created an unusual expression that consumers wouldn’t normally use in everyday language. The combination was therefore considered sufficiently distinctive to function as a trade mark. For a short time, this suggested that simply combining descriptive words in a slightly unconventional way might be enough to secure protection. But the Court soon adopted a stricter approach.
This consumer-centred approach has since become the cornerstone of modern EU trade mark law. Today, courts are less interested in whether a word is technically a new invention and more interested in whether its meaning is immediately obvious to the average person encountering it.
That reasoning explains the outcome in OpenAI v EUIPO. The General Court concluded that most English-speaking consumers would naturally read OPENAI as two familiar terms: “open” and “AI”. In the context of software and artificial intelligence services, those words immediately suggest accessible, transparent, or open artificial intelligence. The fact that they appear as a single word doesn’t change that perception, consumers routinely separate compound words into their individual components and derive their meaning almost instantly.
Perhaps the most important principle the Court confirmed is that a term doesn’t need to have only one possible meaning to be considered descriptive. If one reasonable interpretation describes the goods or services, that’s enough to prevent registration. So even if “open” could mean different things in different contexts, as long as one of those meanings describes the AI-related products and services at issue, the mark cannot be registered.
Reputation Is Not Enough
Perhaps the most revealing aspect of the judgment is its indifference to OpenAI’s fame. The company is globally recognised. ChatGPT has hundreds of millions of users. The “OpenAI” brand features daily in international media. Yet none of this mattered and the reason it didn’t matter exposes a deliberate structural choice within EU trade mark doctrine with significant policy consequences.
The reason lies in the distinction between inherent distinctiveness and acquired distinctiveness. The trade mark system refuses to allow a single early mover to lock up descriptive vocabulary simply by being the first to build a reputation around it. Reputation, in this framework, is a factual state of affairs; distinctiveness is a legal conclusion that must be independently established.
The policy rationale is straightforward: if fame alone could convert descriptive terms into proprietary marks, the first company to achieve dominance in any sector would acquire a linguistic advantage over later entrants. The doctrine forces a separation between commercial success and legal monopoly, a separation that is arguably more important in rapidly evolving sectors like AI than in mature industries where descriptive vocabulary is already well mapped.
Why This Matters for AI Companies
The OpenAI judgment has implications far beyond one company’s branding strategy. Companies routinely incorporate terms such as “Open”, “Smart”, “Auto”, “Vision”, “Generative”, “Deep”, and “AI” into their brand names. When these terms were novel, they might have carried a distinctive aura. But as public understanding of artificial intelligence deepens, the semantic space available for descriptive interpretation widens correspondingly.
This trajectory mirrors historical patterns in other sectors. The practical consequences for the industry are threefold:
- Branding strategies built on descriptive AI vocabulary carry significant and growing registration risk. What feels distinctive today may be unregistrable tomorrow as public understanding evolves.
- Companies that have already built substantial goodwill around descriptive names face the burden of proving acquired distinctiveness under Article 7(3). That’s a high evidential bar requiring extensive survey and market evidence.
- The judgment creates an incentive structure that favours fanciful or coined marks, think “Anthropic”, “Mistral”, or “Cohere”, over transparent, descriptive names. Whether that incentive structure ultimately serves consumers, who arguably benefit from descriptive branding that communicates product characteristics, is a question the Court didn’t address but that the AI industry will need to confront.
Conclusion
The OpenAI judgment isn’t about OpenAI losing. It’s about the evolving boundaries of proprietary language in an era of rapid technological diffusion. The General Court’s decision upholds a principle that has run through EU trade mark law since its inception, one that the doctrinal evolution from BABY-DRY through Postkantoor and Biomild has progressively sharpened: descriptive words belong to everyone, and no single undertaking may claim exclusive rights over vocabulary that the market needs to describe its products and services.
But the judgment also raises questions it doesn’t answer. If the cognitive-transparency test continues to tighten, the range of registrable compound marks in technology sectors will narrow considerably, potentially pushing companies toward arbitrary or invented names that communicate nothing about their products. That outcome protects competitive access to descriptive vocabulary, but at the cost of the informational function that descriptive branding serves for consumers. It’s a trade-off worth watching as the AI industry matures.
