Insights Who owns AI art?


Who owns AI art? It’s a question that’s becoming harder to answer as AI becomes increasingly sophisticated.

Recently, AI art has stormed social media as a Beta version of OpenAI’s software ‘Dall-E2’ was made available to all. DALL-E2 is a machine learning model that generates digital images from text inputted by a person. People have been using their creativity to push the software’s boundaries, with output including interesting transformations to famous paintings such as ‘Mona Lisa with a mohawk’, but also the potential to create vivid and stylistic new works of art via prompts such as ‘an abstract piece showing an elephant playing poker with a tiger in the moonlight.’

These fascinating results raise fresh questions around who owns these new creations, and what it might mean for the future of copyright work created by AI.

At present, OpenAI sidesteps these thorny questions by giving users the contractual right to commercialise the images they create. They don’t, however, grant ownership of copyright over the image. This may work for now as people experiment with the tool, but OpenAI can just as easily take away a creator’s commercialisation rights as quickly as they gave them under the contract.

This uncertainty poses difficult questions for people who may want more security over the content they ‘produce’ with their prompts. It’s not hard to imagine a situation in the near future where people create a unique art style using AI systems, possibly by feeding in their own art portfolios with unique styles, visions and texts prompts, to then sell to paying audiences.

The UK is one of the very few places in the world that offers computer generated work some protection.

Under section 9(3) of the Copyright, Designs and Patents Act 1988 (CDPA), artistic work which is computer generated work (CGW) shall be owned by the person who makes the necessary arrangements for the work to be created. Such definition is, of course, ambiguous.

In the absence of any concrete case law to clarify what is meant by “necessary arrangements”, questions arise about what level of input is needed and by whom. Would the owner be companies such as OpenAI, whose employees created the AI code that artists rely on, or would it be the artist themselves, whose text input was necessary to create the art? Would there be some sort of joint ownership of both? And how does that fit with the ‘work’ being the expression of the authors intellectual creativity – whose creativity is expressed by an AI programme? Often the effect of this depends on the facts and circumstances, with both the company and artist having a valid claim to the work.

In its recent ‘National AI Strategy’, the UK government invited organisations to consult on the interaction between AI and intellectual property. Particularly, if copyright laws around CGWs should change to encourage AI technology adoption and promote its use for the public good. In June 2022, the UK government decided to keep the law as is – meaning that CGWs would have copyright protection for 50 years (s12(7) CDPA), which is less than that granted to work which is fully authored by a human (70 years).

The consultation does little to alleviate concerns around who owns AI art. The grey area between works created using a computer and works created by a computer still clouds the water. The artist using the software who wishes to exploit and protect the images they create, and the person or company who created the programme would both have a legitimate claim to the content.  However, the consultation provided an interesting alternative that may prove useful as the worlds between AI and art become more intertwined.

As part of its options, the UK government suggested that the length of protection offered could be varied depending on the level of effort and/or investment put in by the computer’s efforts. This new form of protection would sit alongside any other rights that subsist in the work. So, it’s possible that this new form could protect both human and AI creativity.

With this in mind, a possible threshold could develop for where ownership lies. Inputting some simple text might not trigger ownership but inputting unique creative visions into the generator could tilt the rights in the artist’s favour and may fit better with protecting the ‘expression of the authors intellectual creativity’ in line with existing case law.

It seems we are some way off this issue being clarified, but there are positives signs that with some proper guidance, the answers to these questions might become clearer.