Five questions creators should ask every AI company

When I worked in news, there was a persistent pressure to write as frequently as possible and push your stories through as many social media channels as you possibly could. Today, content creators need to deliver far more content across a growing number of distribution channels – far more than when I was in a newsroom. Business Insider reports creators feel a growing need to use AI to meet exploding content demand – especially creators earning higher incomes. In a report shared exclusively with BI, researchers found “six-figure creators and artists in the US were more than three times as likely (64%) to say they felt "large" or "very large" pressure to use AI, compared with the lowest earners (19%).” The more money a creator made, the more likely they were to use “fully AI-generated music or content in their work” with those creators earning $50,000 or more “embracing AI at higher rates”.

Here’s the problem: fans hate AI slop – the content that comes from large models trained on data IP owners never consented to have used for training. Most fans bail when they catch even a whiff of slop, moving their attention and the dollars that come with it somewhere else. In some cases, fans may even advocate for others to stop following a creator if they detect AI slop in their content. Celebrity-tier creators and professional studios can’t use most AI models because of IP and security concerns. Regardless, AI isn’t going anywhere, and there is more of it popping up in creator content than less.

So, what should creators do if they want to use AI without the risk of eroding their brand integrity? I worked in both the media and technology industries, which gives me a unique, multi-disciplinary perspective on this challenge. In addition to that, my co-founder, Dr. Sarah Laszlo, and I are building a consent-first AI company called Sublimate, which is designed to serve creatives. We have asked and answered a lot of questions related to AI use in media. Given that, here are the questions I recommend creators ask before incorporating any AI into their workflow:

• Do you own the data the model is trained on? If the model is trained on petabytes of unsourced data, steer clear or at least think twice before using the model to generate content for your audience – especially if you distribute content for money. Inadvertent unauthorized use of protected intellectual property is a real risk with models trained on data with questionable provenance. It’s also an easy way to fall into the slop trap, because models trained on that much data will regress your content to the same mean as everyone else’s. As a creator, you know people don’t want average – they want you! Holding an audience’s trust depends on preserving the ability to deliver what’s uniquely yours to deliver. Anything that regresses to a statistical mean will make that much harder to do.
• Do you know who holds the generation rights to your likeness and IP? A model developer needs two rights: the right to train with your data, and the right to generate against it. Deepfakes are the result of companies not being held to account for identifying and protecting your right to say who is allowed to generate against your likeness and IP. Work with companies that identify your generation rights and say where their right to generate using your content begins and ends in clear terms. I insisted that we outline generation rights in our company because, as a creator, I want to have a say when it comes to who generates using my prior work and likeness.
• What are you using the model to do? Using large models with questionably sourced data is all but unavoidable at this point. I would be a liar if I said I didn’t use LLMs, but I am careful about which ones I use and what I use them for. For example, I don’t like to use them to write posts like this, and I don’t use them to generate synthetic or real people. I use them to sort through large volumes of data, identify spam and scam emails (people use AI to make them, so AI is good at identifying them), and organize information. I also use them to make infographics and animations, building on original content I’ve made. I also use them for coding projects, which is something they are uniquely good at – far better than I will ever be. I do not use them to recreate my likeness, I use our private, secure, and isolated models at Sublimate to do that.
• Is the AI model your asset or someone else’s? Most AI models are assets of the company that trained them. We designed our models to be assets for creators that they can control. We let them control who accesses their model, how much they want to charge third parties for access to the model, and how much they want to charge for generations from the model that a third party creates. We also give creators a full log of activity against their model so they know how their IP was used. This was important to us because if a model is trained on your data, you should be able to leverage it like any other asset you would invest in.
• Was the AI model built with creators or simply built for them? Tech companies haven’t exactly been a job hub for artists and media professionals. These companies have built tools to service creators, but very few artists are offered equity or a seat at the table when it comes to decisions about the direction a technology will take. We made a point, at Sublimate, of bringing artists and creative professionals around the table. Whenever we can, we try to work directly with artists to inform how we build our tools. We do this because artists know best what their opportunities and constraints are. They know where the gaps are in the market of creator tools. If an AI company didn’t have creators there from day one, it’s worth asking if they have your best interest at the heart of their product.

These are just a handful of questions I believe creators should ask before signing up to use any AI model to create content, especially if it features their likeness or the likenesses of other people. Creators have the power to determine which AI companies will live, die, and thrive over the next few years – more power than they may realize. Standing with companies committed to building new tools and assets alongside creators that identify, respect, and preserve their rights to make content authentically theirs will begin paving a new, and I believe, better path for AI.