AI and Trademark Reputation: How ChatGPT Got Famous in 18 Months

A century of continuous use built the reputation of Renault, Hermès and Dior. ChatGPT did it in about eighteen months. Twelve trademark decisions handed down by the French PTO (the INPI) between 2018 and 2025 show artificial intelligence doing two things at once to European trademark law: collapsing the time it takes a mark to become “reputed,” and dissolving the walls between industries that used to have nothing to do with each other. For US counsel, both shifts land in an area where European law already diverges sharply from the Lanham Act.

What “reputation” means in Europe — and why it is not US “fame”

A US practitioner reads reputed mark and reaches for dilution: Lanham Act §43(c), famous marks, blurring and tarnishment. The instinct is right but the threshold is not. US dilution protection is reserved for marks that are “famous” — widely recognized by the general consuming public — and since the 2006 TDRA, niche or sector fame does not qualify. European “reputation” sits lower: a mark is reputed if it is known by a significant part of the public concerned by the goods or services it covers. Sector fame is enough.

Reputation in Europe also unlocks a head of protection the Lanham Act does not have. Beyond dilution-style harm (detriment to distinctiveness or repute), an EU or French owner can block a later mark that would take unfair advantage of the earlier mark — free-riding, riding in the reputed mark’s slipstream. There is no clean US analogue. This matters below, because AI marks are exactly the kind of magnet for free-riding that this ground was built for. For the full mechanics, see our pillar pages on reputed trademarks in France and reputation at the EUIPO.

The speed problem: reputation in eighteen months

On 1 July 2025 the INPI recognized the mark CHATGPT as reputed. OpenAI had launched the service in November 2022. The office found that CHATGPT “designates an artificial-intelligence model used by several million users worldwide and recognized by the general French public,” and that the tacked-on .FR was non-distinctive — it merely signals a national domain. The opposition against ChatGPT.FR succeeded (INPI, 1 July 2025, OP 24-3222).

Point of attention

The classic reputation factors — market share, length and intensity of use, advertising spend — assume years of trading. For AI marks the INPI has accepted reputation built on viral adoption and organic media coverage, without the traditional commercial paper trail. Recognition, not longevity, carried the day.

The same acceleration appears elsewhere. In SALESFORCE / COMPTAFORCE (INPI, 19 June 2025, OP 24-3844), the office found that the shared suffix -FORCE created a mental link with Salesforce even for dissimilar accounting services. The reputation of a modern tech brand did the work that decades of trading once required.

The convergence problem: when a carmaker opposes a smartphone

The more disorienting development is sectoral. In RENAULT / RENO (INPI, 10 February 2025, NL 23-0046 and NL 23-0048), the century-old carmaker obtained cancellation of RENO — Oppo’s smartphone brand — for goods including smartphones, tablets, AI humanoid robots and GPS devices. The office described a “global phenomenon of companies originally tied to the digital and smartphone sector extending into the automotive sector,” because connected and autonomous vehicles need connected devices to function.

This is not a French eccentricity. It tracks the EUIPO’s own recognition of “brand extension” into neighboring markets, where the public will readily assume a natural product-line stretch. The practical effect for a US client filing an AI or connected-device mark in Europe: the prior-rights universe is wider than the class headings suggest. An automotive mark can now stand in the way of a robotics or software application. Convergence was also decisive in the HYUNDAI cancellation, where a third party’s HYUNDAI mark for “AI robots” and “intelligent personal-assistant software” was struck down as an encroachment on the carmaker’s reputation.

Bottom line for clearance

Before filing an AI-adjacent mark in the EU, run the clearance search across tech and automotive, luxury and industrial registers. The principle of specialty — protection limited to the registered goods — bends when reputation and convergence combine.

Where reputation still stops: total dissimilarity

European reputation is powerful, not unlimited — and the exception is instructive. In DIOR / FARDIOR (INPI, 14 April 2025, OP 24-2838), Dior invoked crushing reputation (over €300 million in 2021 revenue, “one bottle of Sauvage sold every three seconds”) against a mark containing the string -DIOR. The INPI rejected the opposition for home appliances and AI software: the goods shared no nature, function or purpose, no diversification by Dior into appliances was shown, and no link would form in the public’s mind “despite the great reputation of the earlier mark.”

Contrast that with Salesforce, where both sides sold B2B services and a link was plausible. The dividing line is not the strength of the reputation but the presence of a mental bridge between the goods. The same logic sank L’Occitane’s opposition to L’ORTIE OCCITANE (INPI, 2 June 2025, OP 24-3011): the shared element OCCITANE was weakly distinctive because it names a region — even a reputed mark cannot monopolize geography.

Practical takeaway

Defending against a reputation-based opposition in Europe, your best evidence is usually about the goods, not the signs: total dissimilarity plus no history of the senior owner diversifying into the contested field. That is a very different playbook from arguing likelihood of confusion on the marks alone.

What US counsel should take away

Two structural shifts, one old constraint. Reputation now accrues fast enough that a two-year-old brand can wield extended protection — so a “new” AI competitor is not automatically a safe distance from an incumbent. Convergence widens the field of marks that can block you, across industries a US class-based search would never surface. But the specialty principle still holds where the goods are genuinely unrelated and the senior owner has never strayed toward them.

The open question is how far convergence runs. If carmakers become AI publishers and AI companies build cars, the similarity-of-goods test — long the backstop that kept unrelated sectors apart — may lose much of its force. For now, the safe course is the unglamorous one: search wide, and build a mark distinctive enough that no incumbent’s slipstream is available to ride.

Key takeaways

• European “reputation” is a lower bar than US “fame” and adds an unfair-advantage (free-riding) ground the Lanham Act lacks.
• AI marks can reach reputation in months (ChatGPT: ~18) on adoption and media coverage, not decades of trading.
• Sectoral convergence (auto / smartphones / AI) widens the prior-rights field — clear across industries, not just your class.
• Reputation still fails against total dissimilarity of goods with no diversification (DIOR / FARDIOR).

Applicable law

France: Art. L. 711-3 and L. 713-3 of the Intellectual Property Code (protection of reputed marks against unfair advantage or detriment). European Union: Art. 8(5) of the EU Trade Mark Regulation. Both extend protection beyond identical or similar goods where a later sign, without due cause, would take unfair advantage of, or harm, the earlier mark’s distinctiveness or repute.

Decisions cited

French decisions are searchable by number on Pappers Justice.