Stable Diffusion and Legal Confusion: Demystifying AI Image Copyright
Think back. Itโs 2011. The seemingly never-ending Harry Potter franchise has just ended, Adeleโs โRolling in the Deepโ is EVERYWHERE, and planking in weird places is still the most popular social media trend.
Are you overloaded with nostalgia, yet?
Well, thereโs one big thing from 2011 that we havenโt mentioned yet, but that definitely dominated the minds of many: the iconic Monkey Selfie.
Now, you might be thinking:
โI do remember that! But what does a decade-old monkey selfie have to do with image copyright?โ
Well. What you might not know is that, in 2015, PETA opted to sue wildlife photographer David John Slater on behalf of Naruto (the monkey) on the grounds of copyright infringement. What followed was a legal storm that rivalled even the most dramatic episodes of ‘Law & Order.’
Ultimately, Slater went on to win the case on the basis that animals canโt legally own copyright, and the image was kept in the public domain. However, just like Naruto’s selfie, the world of AI-generated images has thrown us into a whirlwind of legal confusion.
After all, if a monkey canโt own copyright, can AI? And given the blurring of boundaries, where do businesses and marketers stand when it comes to using AI-generated images?
Letโs find out.
Firstly, Whatโs the Buzz With AI-Generated Images?
Given their rising popularity, itโs likely youโve heard of (if not used!) at least one AI image-generation tool in recent months. And yes; that includes jumping on the social media trend involving Lensa.ai!
At present, two of the most widely used tools are Stable Diffusion and DALL-E; both of which have the ability to create highly detailed and photo-realistic images from simple text prompts.
And, with these tools being so readily available, it hasnโt taken businesses long to consider how using AI-generated images could streamline their activitiesโฆ
- Theyโre Cost-Effective: Because AI image generation tools can create graphics with minimal human input, the cost of these images is next to nothing. These tools eliminate the need for professional equipment and human talent, ultimately allowing companies to allocate their resources more strategically.
- Theyโre Quick to Create: With just a simple text prompt, AI image generation tools can create graphics at an astronomical rate. This rapid creation can be invaluable to businesses that need to produce a high volume of images in a very short time. It can also be a huge benefit when businesses want different variations of the same image.
- Theyโre Customisable, Scalable, and Consistent: In the same way images from graphic designers are customisable, AI-generated images are, too. Visuals can be tailored to brand guidelines with ease. Plus, as a brand grows and evolves, these images can be scaled to meet ever-changing needs with a level of consistency thatโs challenging for graphic artists to achieve.
They Push the Boundaries of Design: AI can produce innovative graphics that push the boundaries of traditional design. The algorithms they rely on can generate artistic concepts that may not have otherwise been explored. This can spark fresh creativity and expand a brandโs visual horizons.
The Benefits Sound Great! Whatโs the Problem?
Well, to explain this, weโre going to have to get a bit technical, so bear with us while we dust off our law booksโฆ
The current copyright law in the UK is the UK Copyright, Designs and Patents Act 1998. This law covers all intellectual property rights, and it gives the creators of literary, dramatic, musical, or artistic work (plus others) the right to control how their materials can be used.
And this is where things start to get a bit sticky for AI-generated images.
Copyright is automatically applied when an individual or company creates a piece of work. However, in order to be protected, the work needs to be original and exhibit a degree of skill, labour, or judgement. Technically, this criteria isnโt met when AI creates new works.
Furthermore, copyright canโt be claimed for any work that copies or uses aspects of an existing work. Even if youโre granted permission from the original owner to use certain elements of their creation, the elements of their work you use to create a new work are still owned by them.
Because of these caveats, without the consent of the original owner, businesses and people alike are committing an offence if they:
- Copy the work of another.
- Rent, lend, or issue copies of the copyrighted work to the public.
- Perform, broadcast, or show the copyrighted work in public.
- Adapt the copyrighted work.
Now, AI image generators have been trained to create images using a database of pre-existing graphics, photographs, paintings, illustrations, and text prompts. They learn how to identify which types of images match certain phrases in order to create results in line with the user query.
However, and this is a biggie, these training databases often include copyrighted or licenced imagery. And thatโs the problem.
Now, you might be thinkingโฆ
โBut, Heinz Did It! Why Canโt We?โ
Well. Before we explain, let’s give an overview of Heinzโs campaign to our less-informed friends.
Heinz gave AI image generators one prompt: Ketchup. The result? Hundreds of images of tomato ketchup that looked exactly like – you guessed it – Heinz ketchup bottles. The campaign was a huge success, going viral on social media and online, and proved that even AI chooses Heinz.
Clever, right?
Sure. Really clever.
But, thereโs a key thing that you need to consider before starting to come up with your own viral AI campaign: Heinz asked AI to create renders of a product it designed and owns.
Itโs this differentiation that makes it highly unlikely the condiment giant will land in legal hot water.
And just to prove the point: Getty Images – one of the worldโs largest image libraries – is currently suing Stability AI, creators of the AI image generation platform Stable Diffusion, for infringing on their intellectual property rights.
Getty Images has been open about its support of AI and regularly allows AI companies to purchase image licences in order to use their assets in AI training databases.
But, Stability AI has applied for no such licence.
Therefore, Getty Images is claiming the AI company has infringed on intellectual property rights including the copyright and licensing of imagery owned and created directly by Getty Images.
Although the case is very much ongoing, itโs further highlighted the need for rules surrounding copyright and AI to be put in place for the protection of all involved.
So, Whatโs Best to Do?
The truth is, the best way to avoid the potential copyright infringement lawsuits that come with AI-generated images is to avoid using them for commercial purposes, if not for everything.
Alternatively, considering many image libraries, such as Getty Images, offer licence packages AI companies can purchase, it may be best to reach out to the owner of your AI image generation tool of choice. Simply ask them if the training images provided to their AI tool were free from copyright, and they should be able to give you a helpful answer.
Aside from this, unless you have your own product (such as Heinz) – where infringements will be less likely – as with all AI, itโs best to approach with caution rather than diving in head first.
On the off chance you arenโt in the business of having your own products, this article from Soap Media offers some great ways you might be able to scratch your commercial AI image itch, without getting in trouble.
So, are you feeling demystified now?
Gareth Hancock has been working as a Graphic Designer for over two decades and is currently Head of Design at digital agency Soap Media. His work includes everything from tongue-in-cheek marketing collateral and fresh rebrands, right through to full-scale website design.