7 Ethical Issues for Creatives Using Generative AI

Using AI for client work can be ethical, but the real questions are related to your response to the brief. The key questions are: Are you meeting the client’s expectations? Are you being honest about what they are paying for? Are you delivering work they can use without the risk of copyright claims? AI can be a useful component of the development process eg using ChatGPT to assist in video preproduction including planning and scriptwriting. Another example would be using Adobe Firefly to generate character ideas and background images for an animated presentation.

Questionable ethical behaviour occurs when AI is used to misrepresent time allocated to the project, provide misleading information about the originality of deliverables or expose the client to legal problems associated with copyright claims. Essentially, cutting corners by significantly reducing reliance on creative skills. It is also worth noting that existing AI tools do not offer a satisfactory ‘one click’ answer to a creative brief when compared to the refinement and sophistication of solutions produced by intelligently integrating AI into workflows.

Disclosing whether generative AI was used (and the level of use) in a creative process is a matter of judgement from an individual or creative team. Disclosure is recommended if the level of use may cause trust issues with the client or there are potential legal risks associated with copyright for example. Where these risks exist it is important to consider using AI generated by reputable companies with trust policies in place such as Adobe.

Effective use of AI tools involves knowledge and experience so, where AI plays a significant role, a creative team may wish to demonstrate to the client how AI has contributed to other projects. It is more respectable to have proactively demonstrated the use of AI and how it provides an edge over competitor’s solutions rather being forced to respond to clients questioning creative methods and processes.

Use of generative AI trained on copyright material is contested because what may be considered to be “legal” is not necessarily “fair”. Where creators did not consent to their work being used for training, claims of copyright infringement may be evident and valid. Other cases are harder to prove – either way many artists feel this undermines their livelihood. As a creative professional ethical risk can be reduced preferring models / tools that offer clearer licensing, opt-out mechanisms or training transparency. Note that creative software industry leaders Adobe have a comprehensive policy related to their training material and models whilst providing solid assurances from a copyright perspective.

It may not always be possible to fully audit training data, but it is possible to control how you use the tool. It could be useful to employ a reverse image search tool such as TinEye or by searching for similar images on Google Lens for additional reassurance. The best policy when considering the use of AI tools in creative media production is not to take unnecessary risks – without a solid licensing policy there is a degree of legal exposure. Where licensing policies are unclear, generative AI can be a useful source of inspiration similar to that of visiting an art gallery or browsing through a graphic design magazine.

Generating AI content in the style of an existing artist can negatively impact your professional reputation, even if it is useful or desirable as part of the creative process. An artist or designers ‘style’ is about reputation, labour and identity which has often taken many years to establish. Using an artist’s name to get their signature look is like extracting value from their brand without permission or compensation. This can discredit you as a creative practitioner. A more appropriate approach would be to objectively describe a styles’ characteristics and allow the AI system to work its magic.

An example might be “high-contrast, monochrome, atmospheric, medium format photograph” (along with any additional prompt information regarding content) rather than “a photograph in the style of Anton Corbjin”. Alternatively, licensing artwork, photography or design work from reputable sources or even hiring the artist can resolve any ethically questionable activity. Remember, if the original training data has used copyright material, even if you prompt ethically there is a risk of significant similarity or ‘likeness’ leaving the creator open to potential legal problems.

Ethical AI systems and their usage consider and recognise that creative ecosystems involve livelihoods. These tools should integrate into creative workflows by enhancing and streamlinin the work of creatives such as designers, video producers, photographers or character animators. They should be used to work smarter and more effectively whilst maintaining creative integrity, providing an edge over those who shun the technology altogether.

As previously discussed in this article, one-click solutions usually result in significant creative compromise. Where limited design input is required, such as the organisation of a simple table in a Word document, a one-click solution may have its use. For creatives, in the same way modern electronic calculators enhanced the mathematicians daily tasks, generative AI provides a tool that can stimulate ideas and streamline creative processes. That said, without careful management to avoid dependence, practitioners run the risk of gradually losing their creative edge.

Obviously, generative AI has an environmental impact simply by existing. For designers, video producers, and photographers, the main issue is not usually one single prompt, but the wider infrastructure behind it. The biggest impact comes from data centres because they use large amounts of electricity, water for cooling and hardware such as GPUs. Smaller jobs such as object removal, AI masking, captioning or generating a few low resolution images are usually less impactful because they rely on inference rather than training a model from scratch. However, repetitive image generation with large batches of variations, long AI video generation, high-resolution upscaling and custom model training create significant load.

There is an alternative view: generative AI can sometimes reduce environmental impact by cutting the travel associated with video production or photography as well as limiting printing or shipping. The counter-argument is that savings are cancelled out if AI leads to excessive experimentation or intensive cloud-based workflows. In conclusion, if creation processes create efficiency, generative AI is neither automatically good nor automatically bad for the environment. In reality generative AI is often a ‘hit-and-miss’ process with large numbers of unused generations.

A good baseline when planning, developing concepts and presenting creative work is honesty about AI content + checking consent in the training data + accountability for use. Be honest about what you did, seek consent where rights and confidentiality are involved, and remain accountable for the final output (including bias, harm, infringement risk, and quality).

Avoid emulating the work from other artists or designers. Do not generate lookalikes of real people without permission. Keep documentation or proof of your process and use licensed/permissioned inputs where possible eg Adobe Firefly. Ensure your input to the project is clear, tangible and measurable. Ultimately, if you feel uncomfortable explaining your workflow to the client, the audience or a peer you respect, that’s a sign you are likely not acting in an ethical manner.