They say imitation is a sincere form of flattery. But if you're a creator, artist or brand, preying on your hard work doesn't feel like something to celebrate.
Small artists have fought the stolen work and used it for many years for commercial purposes, but the advent of AI images and video generation tools has allowed them to abuse copyrighted material more than ever before.
SORA 2 – Copyright Abuse Machine
Following the release of Openai's SORA 2, users were able to create AI videos using copyrighted characters such as South Park, Rick and Morty, and Pokemon. Users were also able to create Deepfake videos (including videos from Openai CEO Sam Altman Shoplifting).
Altman quickly published his blog post (probably to dodge major lawsuits) as it became clear that “training” data was being used in live products.
His solution? If you do not wish to use copyrighted materials, you must expressly opt out or report infringement. Incidentally, there is no apology for any damages that could have been done.
Needless to say, users may be able to circumvent these restrictions if they tweak the prompt. This can be more difficult to remove, as Altman implies.
There may be some edge cases of generations that shouldn't get over it, and it takes some iteration to make our stack work well.
The next part of his response basically suggests that it is actually *good* that a user can create anything they want with copyrighted material.
We've heard from many right-handed sholders who are very excited about this new kind of “interactive fan fiction” and believe that this new kind of engagement will have a lot of value for them, but we want the ability (not included at all) to specify how the character is used. Assume that different people try very different approaches and figure out what works for them. But we want to apply the same standard for everyone and make sure we decide how to go to the right shoulder (Of course, our aim is to make it as attractive as many people would like). There may be some edge cases of generations that shouldn't get over it, and it takes some iteration to make our stack work well.
I personally haven't seen how anyone is creating AI slop videos. PikachuFightingRonald McDonald is particularly thrilling to brands where intellectual property is being featured, and to the fact that the risk of misuse is far outweighing profits. But no matter what slops people create, these brands are so well known and authoritative that they definitely can't do much damage. At least regarding licensed characters (more on later).
However, small creators, artists, studios, brands, especially those who don't have a huge legal team behind them are more likely to infringe copyright, even if they don't know that. Also, in some of these small entities, copyright issues can destroy them overnight, especially as AI becomes more sophisticated.
Imagine you are a promising beauty brand. What stops users from putting one of your products on their skin and then creating AI videos showing that their skin is broken while burning their skin?
Perhaps you are a small chain of family-friendly restaurants. Users can infuriate the restaurant's mascot video and launch a little child across the car park.
And some people believe it *.
And that's where big brands can still catch. Yes, it may be possible to request that you opt out of using your character or other intellectual property licenses, but what about the location, staff uniforms and other brand aspects that make them instantly recognised? As Starbucks, what video do you want to watch? It seems there is Is one of your baristas throwing a hot Americano into someone's face? Be honest – if they said the barista did it, they didn't want to write Charlie Kirk's name in the cup for their customers… how many people would share it with you thinking it was true?
In an age of boycotts, there is also a great risk that “fake news” will be created around brands and related parties. Also, there is little way to convince people that what they see and hear is simply not real.
Part 2 of Altman's blog talks about monetization. But primarily for Openai – not creators or brands whose works have already been plagiarized.
Secondly, we need to make some form of money for video generation. People are generated far more than expected per user, and more videos are generated for very small audiences. Try sharing a portion of this revenue with the right sholders that users want to generate characters. Accurate models require trial and error, but we plan to start soon. Our hope is that a new kind of engagement is even more valuable than revenue share, but of course we both hope that it is worth it.
At the time of writing, Sora was only invited and apparently number one on the Apple App Store charts on iOS only. If Openai admits that you need to make money from video generation because it is already in high demand, if you need to make revenue share, how exactly do you expect users to charge you to pay for this when it becomes more widely available?
Please pay attention to the parts.”Our hope is that a new kind of engagement is even more valuable than revenue share. This type of person slaps the type of people who ask the artist to work for what they want to increase their exposure. Can I pay my bills by exposing it? No, I didn't think so.
Overall, I don't think they will do so unless Openai and other AI image and video generators are forced to seriously tighten copyright misuse due to huge class action or government intervention.
At the moment, I think there are plenty of uneasy marketing, legal and public relations teams, but it is set to become increasingly difficult to monitor the brand's reputation and emotions, as it is likely that Sora's copycat will emerge in the coming months.
