KMK Law Cannabis & Craft Beer Blog

In response to a rise in craft beer trademark disputes, research scientist Janelle Shane recently set loose a "neural network" – a type of computer program that can "learn" from data sets that are given to it – to create new craft beer names. The thought was that the computer program could learn from thousands of existing beer names, and come up with a long list of new ones built on detected patterns. Last month, Shane published a list of dozens of names, saying: "It worked . . . I give you: craft beer names, invented by neural network."

Is this a solution to the craft beer trademark problem? Maybe it is a fun and helpful start, but I would say: not really. The problem is that trademark rights are not an exact science based just on whether you've technically found a unique combination of words. The test for trademark infringement is whether there is a "likelihood of confusion" among consumers, and it does not require an exact match between products' names.

Some quick searching created doubt as to whether the neural network's beer names are really free and clear of the trademark problem Shane set out to solve. For example:

• The neural network came up with "The Moon" stout. But the "Moon Brewery" already exists and makes beer. It seems plausible that a consumer would think that "The Moon" stout comes from Moon Brewery. Some beer drinkers might even think it is a dark offshoot of the well-known Blue Moon Belgian wheat.
• The network also coined "Third Danger" strong pale ale. But "Danger Ale" is already on the market, from Castle Danger Brewery. In the context of a bar or a beer aisle, it doesn't seem far-fetched for a consumer to think that "Third Danger" is one more in a series of "Danger Ale" from the same source.

A product or package design can also take a unique name and plunge it into an arguably confusing context. For example, West Sixth Brewing got in a serious tussle with Magic Hat when the "6" in its logo was in a circle with a star, which Magic Hat argued was confusingly similar to its "Magic Hat #9" label.

And as this blog has noted previously, some courts have found that the beer industry is a context that more easily breeds confusion than other goods.

The point is that reaching a technically unique combination of words is only a starting point. We can thank the neural network for helping with that initial part of the process. But creating and protecting trademark rights in a new product requires thoughtful searching and strategy beyond the initial words. Depending on the projected sales, a professional trademark search might be a worthwhile investment. And until the AI learns some more subtleties about what might engender confusion in consumers, we have probably not gotten past the need for a thinking human to be at the wheel.