Double trouble: IP disputes over pals, twins and copies

Notwithstanding these instances from the United States, Japan is the exception rather than the rule when it comes to the statutory recognition of software patents, meaning the current litigation's outcome is unlikely to have a strong influence on patent office decisions or juridical interpretations elsewhere. That said, in having the world's third-largest video games market, Japan represents a major and growing producer and consumer of interactive media, where any chilling effect experienced by the domestic industry is prone to extend internationally.

Quarreling twins: trademark lawsuit between Gemini AIs

As one of IT's most influential companies and a dominating market presence, Google is no stranger to controversy. To wit, France's Autorité de la concurrence (Competition Authority) fined the technology company €500 million in July 2021 and a further €250 million in March of this year for disadvantaging press agencies on its platforms.

Both penalties concerned breaches of measures designed to create a fairer information economy, although the more recent also implicated the undisclosed use of copyrighted works to train Google's AI. Bard, as the model was previously known, and another system, Duet, were merged in February 2024 to create Gemini. However, the change of name and branding has not distanced the AI from allegations that it impinges on others' IP rights.

On September 11, 2024, fellow California-headquartered business Gemini Data filed suit against Google for trademark infringement. Appearing before the United States District Court for the Northern District Of California, the lawsuit claims that Google exploited its status as a "corporate giant wielding overwhelming power" to "knowingly and willfully" infringe on the smaller company's trademarks.

Founded in 2013, Gemini Data provides a cloud-based platform that allows users to gather and process data from disparate sources. Businesses can then query its AI in natural language to gain insights or extract conclusions. Crucially, Gemini Data holds registered marks in the United States in Nice classes 09 – recorded computer software for use in business enterprise big data collection, transmission, visualization, integration, analysis, management, and storage using machine learning and artificial intelligence – and 42 – software as a service (SaaS) for the same purposes.

The likelihood of confusion resulting from these prior marks was one reason why Google's trademark applications for "Gemini" were refused by the United States Patent and Trademark Office (USPTO) in May and August this year.

"Shortly after the [May 9] USPTO refusal was issued, Gemini Data was contacted by an 'anonymous' entity regarding the possibility of acquiring rights to the 'GEMINI' brand," continues the lawsuit, with the plaintiff coming to the conclusion that the purchase offer was made on behalf of Google.

Gemini Data is seeking an injunction against Google's use of the "Gemini" name as well as monetary awards, including punitive and exemplary damages. Thus, if the lawsuit is successful, Google will need to rebrand its AI offering a second time.

AI vs. artistry: Germany considers copyrights

What do a celebrated photographer and a massive AI training dataset have in common? Not much, unless you are Robert Kneschke, a German stock photographer whose images found their way into LAION 5B, the information storehouse powering AI models like Stable Diffusion. Kneschke was not exactly thrilled when he discovered his carefully crafted photos were being fed to algorithms without his permission.

So, naturally, it was lawsuit time. Filed in Hamburg's Regional Court, Kneschke claimed that LAION e.V., the non-profit behind the dataset, violated his copyrights. In the creative world, using someone's work without asking is akin to taking someone's car for a spin because you liked the color. You might have a great time, but the owner will not be pleased.

LAION's defense? They said it is all in the name of science, citing an exception in German copyright law that allows text and data mining for research. Sounds good in theory, but the reality gets a bit murky when you throw in collaborations with commercial AI companies like Stability AI. Suddenly, the non-profit's argument was beginning to look shaky. LAION argued that Kneschke's images were not locked down tight enough with "machine-readable" copyright protections. According to them, only a super-technical blockade, like a robot.txt file, could have stopped their hungry AI scrapers. Kneschke, on the other hand, said his stock photo agency had already prohibited automated data collection in its terms — more like a polite but firm "no trespassing" sign, which LAION apparently chose to ignore.

On September 27, the court ultimately sided with the defendant. In a landmark decision, the Hamburg Regional Court ruled that LAION's use of Kneschke's photos fell within the legal bounds of Germany's text and data mining exception. Since the dataset was used for scientific ends and there was no proof of a decisive commercial influence on the research organization, AI remains free to digest creative works.