Our message on Artificial Intelligence
last updated 28 July 2023
We recently sent our artist’s a survey regarding the rise of Artificial Intelligence (AI), and we want to thank them for taking the time to share their thoughts and concerns with us. It was enormously helpful to have those honest and open thoughts. We have collated and processed carefully the feedback, and are committed to addressing the questions, concerns, fears and curiosity to the best of our ability. Here is our message we shared with them:
We are sure that the latest developing news around Artificial Intelligence and how it can create artwork is concerning for you all, but we want to assure you that we are already deeply involved in various initiatives, networks, and newly formed associations in order to find ways to protect you.
At The Bright Agency we are adapting a two-pronged approach to AI and its impact on the industry. Due to the speed at which this new technology is developing, and with so many unknowns, we accept that we need to be both protective and explorative at the same time.
Legal Protection
We are taking steps to ensure that your artwork is protected and that no one is able to use your hard work to train AI models. To this end we are closely following legal developments around the globe and we are also working with industry experts on technological measures we can take to protect your artwork online.
Moving forward, we will provide a series of updates, facts and opinions on this topic for you to digest and consider.
Creative Exploration
With any new technological development there will come opportunities and we are exploring ways in which our artists can utilise AI to help in their creative process. We want to make it absolutely clear that we will never advocate for our artists to use AI to generate their work. However, should they wish to explore the benefits of the technology, legally and ethically, then we will support them in doing so.
This is a developing issue for our industry but rest assured that The Bright Agency is entrenched in many associations, influencing government policy and collaborating with a wide network of organisations, all working towards the same goal – to represent and support the interests of creators, to protect not only you, our artists, but the entire creative community.
You can continue to do what you do best and create great art, knowing that we will ensure you are paid, protected and informed. To this end, here are some of the key actions that we are already undertaking, with more to come.
Monitoring the Getty Images vs. Stable Diffusion lawsuit
In the fast-evolving field of AI-generated artwork and all the legal issues that arise from it, we are following the lawsuit filed by Getty Images against the Stable Diffusion AI image generation system. In this lawsuit, Getty has accused Stable of misusing more than 12 million Getty images to train its system.
This lawsuit is one of three lawsuits currently filed against Stable Diffusion, including a separate case filed by Getty in the UK and a class action complaint filed by artists in California.
This is an ongoing case and will set a precedent for how AI companies are allowed to use copyrighted images for the purposes of training AI systems. As this is the first high profile case of its kind it will undoubtedly also affect changes to copyright laws and regulations.
There has been a call from leading figures in the tech world for a six-month moratorium on further AI development until different industries get a better understanding of the technology. The message to everyone is to not use AI for final/finished artwork as we do not know where the images are coming from.
Notifying Bright if you see or suspect your artwork has been used for AI training purposes
If you see or suspect that your artwork has been scraped for the purposes of training an AI programme, then notify your Agent immediately. We can issue takedown notices to the AI creator and/or the platform to remove your artwork entirely from the site. But bear in mind, until there are concrete changes to the legal framework – hence why the Getty vs Stable Diffusion lawsuit is so crucial – there is little action we can take against individuals who misuse artwork. The law simply hasn’t caught up with the technology yet.
If, meanwhile, you see that a company has used your artwork then let us know and we can send a cease and desist notice and follow our standard unauthorised use of artwork policy.
Implementing legal language and anti-AI provision on client contracts
We are working closely with the SAA and have taken their advice on legal language to include in all client contracts going forward, whereby clients will have a legal obligation to take best measures on their websites to prevent any data mining of artwork and to not use artists’ artwork to train any AI programmes themselves. We have been implementing this for the past month and thus far all clients have committed to this obligation.
Using cloaking technology and protecting your artwork online
As AI develops and more and more artists and businesses are concerned by what its implications could be, there will be more efforts made to prevent AI from scraping artwork. Organisations and universities are already researching and developing software to prevent artwork being copied for the purposes of AI training. One of the more recent studies is the Glaze Project - developed by the University of Chicago, which uses cloaking technology. This is free at the moment and is being used for research purposes but there will undoubtedly be more software like this developed in the future.
We are currently advising our artists to use this on their personal online portfolios and are investigating how we can implement something like this on Sparklebox. We will continue to monitor and research this type of tech as it develops. We also advise artists to put watermarks on their images so that when an artwork is scraped the tech also scrapes the watermark, thus making it easier to identify while also making the AI-generated image worthless to the end user.
Non-artists vs. artists using AI
There should be a clear definition of what constitutes solely AI-generated work and work made with the intervention of creators.
Artists: illustrators, designers, creatives that are commissioned to create an image/artwork.
Non-artists: coders, users, technicians, editors, publishers, production managers.
AI-generated artwork: The non-artist has generated the artwork via text, using subject matter, style and artists’ names to generate a final image.
We support the Copyright Alliance that states:
Creators should be protected and recognised for the copyright and moral rights that exist in their work. Copyright is given to works of human originality and skill and labour. AI-generated works with no human input should not attract copyright protection. Which means that images that have not been created by a human cannot be used commercially.
AI-enhanced artwork: When an artist/designer uses AI as part of their process but the final artwork is not generated solely from the text but rather by the artist using the AI tool as part of their process. These images can be copyrighted and used. The benefits of AI, of speed or reference, have been employed by a human and an artist, using the tool.
For those artists that are interested in using AI to speed up their processes, we will share how we see some artists use the tech and what pitfalls to look out for.
By encouraging artists to explore AI technology as a tool, we are not supporting AI-generated work. However, we do think artists that are informed on all parts of AI – the benefits, the negatives, how to protect themselves and how to manage the powerful emotional responses to it – will only be empowered to make the right choices for themselves.
Our interest is that you continue to be commissioned for the incredible work you do. And that all commissioned creative work is commissioned for an artist to make.
To find out more about the Copywriters Alliance, click here.
Who owns the copyright on AI-generated images?
The question of copyright ownership for AI-generated images is a complex and evolving legal issue. In general, copyright law grants protection to original creative works that are fixed in a tangible form. However, when it comes to AI-generated content, determining copyright ownership can be challenging due to the involvement of artificial intelligence algorithms in the creative process.
In most jurisdictions, copyright is typically attributed to the human creator of a work. This principle follows the idea that copyright protection is intended to reward human creativity and labour. However, as AI technology advances, there are debates and discussions around whether AI-generated content can be eligible for copyright protection and who should be considered the creator or owner of such works.
Different countries may have different legal frameworks and interpretations regarding AI-generated content and copyright ownership. Some countries may consider the human programmer or user of the AI system as the creator or owner of the output, while others may view AI-generated works as falling into the public domain or belonging to no one.
As a commercial illustrator making solely AI-generated artwork, worldwide rights will be tricky to grant if laws on copyright are not unanimous.
If you have a specific context or situation in mind, talk to your Agent to get accurate and up-to-date information regarding copyright ownership of AI-generated images.
There is a global frustration that we are working out the legalities of usage whilst the technology is already open source and available. It’s really not how anyone likes to operate. As we discussed earlier, there are three high-profile legal cases that are currently being observed and which may influence new policies. On a positive note, we have seen various tech companies react positively once they receive feedback.
How can I stop my art being used for training without my consent?
We are fighting for an opt-in model, which means you will get a percentage of however much art you put into the machine, or an opt out. Also, as things currently stand, with no way of copyrighting the image for sole use it can’t be used commercially until this is decided.
We are also exploring a technology called ‘Glaze’. This can trace and track artwork, preventing it from being used to train AI by coating it with a disruptive glaze.
But preventing your art entirely from being used to train AI is a difficult task, as it may be challenging to control or restrict the use of digital content once it is shared online. However, here are some steps you can take to minimise the likelihood of your art being used for AI training:
Watermark and modify images:
Adding visible watermarks or modifying your artwork in a way that does not detract from its aesthetic appeal can discourage unauthorised use. Watermarks can include your signature, logo or any other identifying mark that signifies ownership.
Monitor and track your art:
Regularly conduct reverse image searches using tools like Google Images or TinEye to monitor where your art is being used online. This can help you identify instances of unauthorised use.
Terms of use and licensing:
On all contracts we now clearly define the terms of use for your artwork. This allows you to maintain control over how your art is utilised and ensures that others require permission before using it for AI training or any other purpose.
Educate and raise awareness:
Engage with artist communities, social media groups and online platforms to raise awareness about the importance of respecting artists’ rights. Encourage discussions on ethical AI use and advocate for proper attribution and respect for creators.
With the above, the agency and your agent will always work with you and support you.
The ability to influence our industry’s future is important, as is sharing the unfairness we see and the consequences – for example, we have seen AI providers remove artists’ names from searches.
There is obviously a lack of understanding between each industry, but we have decided that the best way to make a difference and to influence outcomes is by using communication methods that encourage empathy.
We will take appropriate action if you discover your art is being used without authorisation.
What we will do to support you if you see your art is being used by another artist
-
We will begin by issuing a takedown notice to the offending artist. In most cases, when it pertains to an individual acting on their own, a takedown notice is enough. We will monitor their social media and websites to ensure that they have removed the stolen artwork. If they haven’t, we will follow up again until they do.
-
If the stolen artwork has already been licensed or sold to a client, then we will also notify the client so that their legal teams can follow up as well.
If your art is being used commercially without authorisation, there are several scenarios and ways we will deal with this:
-
If one of our clients is using artwork in a manner that has not been agreed in a contract or approved by the artist in writing in any way, then we would notify them of their unauthorised use and reach a settlement agreement with them whereby they compensate the artist financially for the unauthorised use. Depending on the nature of the offense, we will either terminate the agreement with them whereby the rights to the artwork revert to the artist or for an agreed fee we will amend the agreement to include the unauthorised use as being part of the deal going forward.
-
If it’s being used commercially by a legitimate company operating out of a territory with strong copyright and IP protection laws – e.g. US, UK, EU, AUS/NZ etc. – then we will issue a takedown notice and ask them to stop all sales and manufacturing of the product that utilised the unauthorised artwork. In cases where the artwork in question has already been licensed or sold to another company we will also notify them so that their legal teams can get involved as it’s in their best interests to protect their investment in the artwork too.
-
If it’s being used by a clear scam company operating out of a territory with lax copyright and IP protection laws – unfortunately in these scenarios there’s not much we can do as the legal protection isn’t available to us. In most cases, trying to report these companies can lead to further scams whereby they can phish sensitive information about the artist and/or Bright. At best we can make posts in our professional and social media channels, warning our partners about these companies and advising them not to confuse these usages with the legitimate use of the artwork.
To give a deeper insight into all things AI, we’ve provided the following breakdown.
What is generated AI?
Given an input:
Text, emails, social media content, images, voice recordings, programme code and structured data.
It creates an output: Stories, images, blog posts, programme code, songs, poetry, instructions, jokes, music and video images.
It works like a human brain, making billions of mathematical calculations on the probability of where a nose is, how long a leg is, what a stream looks like, for example.
• It’s a huge neural network
• The truth is, they don’t fully know what it’s doing
• Every image starts with billions of pixels and it effectively removes noise from images.
• It’s much more sophisticated than the noise removal slider in other image editors because it actually has an understanding of what the world looks like and an understanding of written language, and it leverages these to guide the process.
It doesn’t have a database of images to reference, it doesn’t use any image processing algorithms at all. It’s all just pure, complex maths.
This example is just a tiny fraction of the logic used during the process. Imagine this times a 1000.
So, imagine pages and pages and pages of this kind of maths. Its equations are designed to process data in a way that is inspired by the human brain. It is a type of machine learning process, called deep learning, that uses interconnected nodes or neurons in a layered structure that resembles the human brain.
How do you access AI?
These are the main platforms: DALL.E 2 (or just DALL.E) This was closed source (no one can see the source code) and up until September 2022 the API was available to only a few. You can now use its API on a pay-per-image basis. Very recently, there was news of Microsoft doing a deal with them to start integrating it into more of its products. Adobe and Microsoft have collaborated in the past so integrating AI with Adobe seems a very likely outcome.
Imagen
Google’s May 2022 version of a text-to-image diffusion model, which is not available to the public (due to ethical considerations regarding the scraped content that trained it). Now that Microsoft has just invested and closed deals with open AI, Google may change their strategy to avoid being left behind.
Stable Diffusion
The most accessible platform due to lots of free online versions and because it’s relatively easy to install on computer versions.
• Released October 2022, v1.5, open source
• Trained on >billion English image-text pairs as well as others (so very English-specific)
• Took 150,000 GPU-hours to train at a cost of $600k initially
• Lots of artists and styles in version 1 redacted in diffusion 2
Midjourney
Currently, only Midjourney allows you to copyright the artwork you generate, and this is on a paid subscription. It also gives a license back to Midjourney and it’s not clear how they would use the image – commercially or just to retrain their AI machine. Either way, this won’t give any commercially used image the protection a client will require.
Midjourney is an independent research lab. It’s closed source and they are working on a web app. Midjourney was in the press recently as it was used to create the illustrations for the book Alice and Sparkle in which the text was also created with AI. This resulted in a discussion around the ethics of AI-generated content.
Our commitment to you is to always prioritise your creativity, uniqueness, and originality. If you need to reach out with any questions, either talk to your agent or contact us here. We are here for you.