Welcome to day 6! Today we are exploring questions and issues involving AI- generated art and content and copyright.

Today’s post was written by Andrew Gray, Copyright Librarian at UAL.

The emergence of AI has raised key copyright questions relating to the role of creativity and how it is recognised. Technology companies such as Microsoft and Google and AI platforms such as OpenAI, Stability AI and Midjourney face legal challenges from educators, artists, publishers and content creators over their AI model training practices.

Training AI on existing works

One of the key legal questions has been:

When you train AI models on copyrighted works, is such training classed as unlawful copying or is it protected under fair use/dealing exceptions?

‘Fair use’ and ‘fair dealing’ are defined as:

Fair Use is used in the United States and allows you to use copyrighted works if it is not substantial, if it is non-commercial, if it ‘transforms’ the work into something new, and if it damages the market or value of the original work or ‘artist’.
Fair Dealing is in the United Kingdom and allows use of a copyrighted work for specific reasons such as research, private study, criticism and review and news reporting. It must not take a substantial amount or affect the market for the original work.

AI generating new works

Another challenge concerns the copyright status of AI-generated works themselves, which raises various questions.

Can AI outputs be protected by copyright?

What constitutes a ‘derivative’ work versus a new creation?

Who owns the rights to AI-generated content?

Different countries, different approaches

These challenges become more complex as different countries take different approaches.

United States has rejected copyright applications for AI-generated works maintaining that copyright protection requires substantial human creativity

China has granted copyright protection to AI-generated content

United Kingdom theoretically allows copyright for computer-generated works but ownership of this remains unclear.

Activities

We know these are big questions and considerations to explore. Choose one of the activities to explore and reflect on how AI tools are being used. Activity 1 involves creating something yourself and activity 2 involves reflecting on images already created. Each activity has a different set of discussion prompts to respond to on Teams.

Activity 1:

Login or create an account with an image generator such as Stable Diffusion. If you are new to using Stable Diffusion, a free account will provide you with 10 “credits” to create AI generations with.

You might also want to use Midjourney (paid) or Adobe Firefly (free with Adobe subscription). UAL staff and students (or others with an institutional enterprise licence) can also use Microsoft Copilot which has a “protected mode” in which your chats and data are not stored or shared with anyone, including Microsoft.

Join us in the Teams space to share your responses and any images you’ve generated. If you don’t have access to the space, email us at teachingexchange@arts.ac.uk for the attention of Hannah.

Using prompts either

Recreate a ‘famous artwork’ but change one major element. For example, the Mona Lisa but make her sad, or The Great Wave off Kanagawa but make the waves a different colour

Create an image/person/object/scene in the style of a famous artist. For example, the Taj Mahal painted in the style of Kehinde Wiley. Here are two examples of this prompt generated using Stable Diffusion.

The Taj Mahal in the style of Kehinde Wiley. In this version, the colours are a little flatter and brighter. There is a blurry reflection of the palace in the water's reflection.

In this version, there are more shadows indicating a different time of day. There is also a clear reflection of the palace in the water in front.

Discussion for Activity 1:

To qualify for copyright, a work needs to be ‘original’, it doesn’t mean it needs to be ‘good’ (luckily for some of us). A recent case clarified the definition of ‘original’ in copyright to mean:

“What is required is that the author was able to express their creative abilities in the production of the work by making free and creative choices so as to stamp the work created with their personal touch. “

Keeping that in mind –

What elements of the original artwork remain? What has changed?

Do you think your new ‘artwork’ could be considered copyright infringement?

Who could claim copyright ownership of this image?

Activity 2:

Each of the following works were created using AI tools.

An AI generated group of people in robes standing in front inside a large room in front of a large round window revealing a mountainside town. The sunlight is causing the room to be back lit.

‘Theatre D’Opera Spatial’ by Jason Allen won an award which was then withdrawn after it came to light that it was created using AI.

An AI generated portrait of young Chinese woman. She is wearing a white robe with a single blue stripe along one side of the collar. Her hair is in one long bride down one side of her face.

’Spring Breeze Brings Tenderness’ by Li was granted copyright protection after someone else took the image and passed it off as their own.

An AI generated image. Half of it has a clearer image of half of a man's face though it's in an unusual combination of pixelated watercolours. The other half of his face is shrouded in a shadow, revealing only the eye and eyebrow.

‘AI God’ was created by an AI robot appropriately named Ai-Da Robot based on Alan Turing and sold for over £830,000 ($1,000,000).

Discussion for Activity 2:

Join us in the Teams space to share your observations and reflections. If you don’t have access to the space, email us at teachingexchange@arts.ac.uk for the attention of Hannah.