AI is a threat to brands. A Brand Corpus can save yours.
One of us was an IP lawyer and one of us was a brand strategist. Here's how brands will thrive in the AI era.
A new threat to brands
If you run a brand, and it’s your job to protect what makes that brand special and to grow it, you are now facing a daunting situation: generative AI has given consumers the ability to do anything they can dream up with your brand’s IP.
This threat is brand new, and it’s more dangerous and complex than anything we’ve seen before.
Brands have often been slow to react during periods of technological innovation to find novel ways to protect their IP. Remember Napster?
Well, this time is different. AI represents the largest loss of control, on the sharpest timeline, that brands have ever faced.
Generative AI models are a new type of threat to brands’ intellectual property (IP) rights. Not only is some of the information ingested by these models untrustworthy source material, like shaky internet sources and old books, but what they spit out can also be misleading, outdated or untrue.
These models are gaining access to all of the information on the internet, making it very easy to create derivative work using a brand’s IP without the brand’s permission. These creations might give the world false, confusing or just plain bad perceptions of the brand.
A lot of these creations infringe on copyrights and trademarks. But while there are legal protections for copyright owners like brands, enforcing those protections against generative AI creations can be difficult due to the complexity of tracking down those responsible and the evolving nature of AI-generated content.
The solution is to build your own Brand Corpus
It’s a chaotic new landscape, but the same generative AI tech will open up new possibilities for brands to derive value from their IP.
Any generative AI tool, whether it’s a text-to-image service like Midjourney or a large language model like GPT-4, is trained on a huge set of data called a corpus. This corpus is built on publicly-available data from the web and books.
This means that models like ChatGPT and Midjourney are generic. They know nothing about your brand that isn’t public, and everything about your brand that is. You have no control over this corpus and how it’s used.
Of course, brands have proprietary data - and lots of it. Archives, internal communications, brand guidelines, marketing copy, marketing images, customer data, sales data, performance data, training documents and even Slack messages.
This proprietary, private data can truly capture the essence of your brand in ways that the public, generic models cannot, and it’s this proprietary data that will be invaluable in the near future.
Meticulously collecting, processing and cataloging this proprietary data into what we call a Brand Corpus is the key to unlocking new value from your IP in the AI era. It should be the focus of every CEO, CMO, CTO and CDO today.
A Brand Corpus is a dataset that contains a brand’s proprietary data and is capable of interacting with both internet-scale data collected by generic AI models and third-party applications, ensuring your brand is maintained and protected across the internet.
Think of a Brand Corpus as a hub made up of your proprietary data - text, images, video, audio, presentations, databases, and more - that serves as the operating system for your brand to unlock AI applications.
When this Brand Corpus is used to fine-tune models like GPT-4 or Midjourney, it allows brands to control and protect their brand as AI corrodes others. Plus, it unlocks the full spectrum of generative AI’s use cases.
Here are a few of the AI use cases we’ve discovered that a Brand Corpus can unlock:
IP and data monetization: Licensing your Brand Corpus for use by other AI models and to other external collaborators for use in new creations
Marketing performance: Arming internal performance marketing and digital marketing teams with a superintelligent analyst sitting with them side-by-side
Customer experience: Supercharging sales teams’ abilities to deliver emotional and effective customer experiences in-store and online
Brand consistency: Effortlessly extending the tone of voice of your brand to new hires, agency partners and all internal members of the organization
Customer retention: Personalize communications with churned customers with performance optimized and custom-tailored messaging and product recommendations
The threat is here
Brands and AI are on a collision course. The explosion of faster, cheaper and better generative AI models in the last year has led to advances in creativity and efficiency that most brands can barely wrap their heads around. We are already living in a world where anyone can use image or text generation to steal your brand’s essence and remix it into creations of their own.
For instance, take the unofficial Balenciaga x Harry Potter mashup that hit Twitter last month, courtesy of AI. It’s incredible, but if you’re more a Star Wars fan, you can check out that version here. Things are officially getting weird, folks.
Adobe Chief Strategy Officer Scott Belsky illustrated an important point in a recent interview:
“I seriously believe [making your own sequels to popular movies and TV shows] will all be achievable very soon and it’s just going to be a copyright dumpster fire. What is going to happen? There’s no way they’re going to be able to commercially monetize that content, there’s just no way, you’re using Spider-Man, you’re using The Matrix, you’re using whatever. If you believe that that’s an inevitability, which I do, then I think that there will have to be clean models that you can rely on commercially.”
Balenciaga did not approve these images, of course, but the far more concerning implication is the fact that millions of users could start slowly eroding any brand’s ability to control its primary source of value - its symbolism, its meaning, its essence.
![](https://substackcdn.com/image/fetch/w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fdc820cf2-9ebd-44c6-9dd3-9c665493a366_1150x657.jpeg)
Brands’ options are limited
You could go scorched earth 🔥
If you’re a brand, you could go after all infringers and infringing platforms and sue them into oblivion.
Some people choose violence. If you’re a brand that wants to go down this road, it’s certainly within your rights. Historically speaking, its probably not going to work out very well for you. One fun but impossibly stupid prediction we’ve heard is that we will simply use AI to streamline lawsuits, set up lawyer bots, and let hundreds of millions of lawsuits sort themselves out in the background. One day, you’ll just get a bill that says you were sued, you lost, and now you owe the aggrieved party $10,000,000, plus interest. No thanks to that.
Brands are no stranger to the Innovator’s Dilemma. Our guess is that c-suites will begrudgingly dismiss the scorched earth option as a mere band-aid that will set them back in the long-term.
You could YOLO 🤪
Give up on policing infringing work entirely and release your brand’s data publicly.
Ben Thompson recently raised the argument that “most of the Internet, given the zero marginal cost of distribution, ought already be considered fake.” Thompson goes on to propose that “the solution isn’t to try to eliminate that content, but rather to find ways to verify that which is still authentic.”
It feels like Thompson is on to something here. He’s suggesting that while the IP levy has broken down, brands would be silly to try and police an exponentially growing infringement source that compounds in power over time. Instead, the right approach might be more in line with the blue check on Instagram.
People want what is authentic. Fans want canon. The only source that can be seen as “official” is the brand itself. It seems much more plausible that brands would find ways to capture the value of its IP this way.
Or you could…
Build your foundation and thrive
Neither of the options is perfect, which highlights the delicate, but necessary exercise brands must undergo. Faced with either scorched earth or DGAF, the smart move is to build an official corpus of your brand’s essence and protect it, license it and nurture it.
Brands have libraries of their own IP from which to draw and they would be wise to start building their own datasets to train AI platforms. In the YOLO option above, the brand goes through this arduous exercise and then…just gives it away.
The leading brands five and ten years from now will be the brands who decided to embrace this moment and defend themselves. It’s a scary time, but a chance to build a legacy. Processing and cataloging IP into a Brand Corpus that’s ready to interface with AI models on the brand’s terms is the only way to lay the foundation for a brand that protects itself against these threats in the future. Starting now will give brands a major head start on surviving all of the exciting and terrifying things to come.
〰️
A special thank you to Brian Mark for editing this piece.
Thank you for putting this together team! We truly live in interesting times and we can't sit idly watching the pitch go by.