Copyright and Generative AI

Introduction

GenAI has brought in a new era in artificial intelligence. With these advances, we have created an ownership issue. GenAI is capable of producing human level artwork, literatur, music, and more for a fraction of the effort.

Copyright law was designed to protect the works of creators. But who is the author of an AI-generated work? Is it the creator of the AI program, the user who inputs data into the AI, or the AI itself? Is AI-generated work original or just a remix of existing work? Courts are left to interpret traditional laws in new contexts.

AI-generated works often involve remixing existing human-created content. This process challenges the copyright criterion of originality, which requires a work to be created through human intellect and creativity.

Can freely available copyrighted materials be used to train an AI? How is this different than a human using freely available copyrighted materials to improve their skills?

If previous creators had not made their work freely available, then Generative AI would not have had any material to train on. Many creators now want to exclude their content from being used to train AI.

The creation of deepfakes presents significant copyright issues. The question arises about the rights to one’s image and the implications for copyright when AI creates virtually indistinguishable copies or alterations of copyrighted material. This became an issue in the recent actors strike.

If a creator takes an AI generated piece and provides human input or process, then is it now copyrightable because it is no longer purely created by a machine?

We need a reevaluation of how creative works are defined and protected in the digital age.

Getting Paid

Should the AI companies pay for each copy of every book, image, or song that is used in the training data? Given that the AI company no longer need that book/image/song after training, could they just loan them instead? Some companies are licensing content for training their models.

It is not impossible to determine which training data contributed to a specific output. Do you therefore compensate every single piece of training data equally?

It is possible to imagine a scenario where a creator loses income because of Generative AI. It is likely that the AI was trained in part on their work. However, the AI is trained on so much more than just their work.

Current copyright law generally requires a work to be created by a human and to be original for it to be protected. AI often creates content by learning from existing copyrighted material. This process can be seen as creating derivative works.

There have been disputes involving AI-generated art, where questions arise about whether such art can be copyrighted, and if so, who holds the rights. In the music industry, questions about copyright for AI-composed music are becoming increasingly relevant.

There is a lack of legal clarity on what constitutes authorship and originality in the context of AI. Legal frameworks have yet to fully address the ethical implications of AI in creative processes, such as the potential diminishing of human creativity and the economic impact on human artists.

The Creator’s Dilemma

Many creators view AI as a competitive force that could devalue their work or make certain creative jobs obsolete. Some creators see AI as a tool that can enhance creativity.

Some creators express concern over the authenticity of AI-generated works, questioning whether it lacks the emotional depth of human-created work. Others embrace AI for its ability to push the boundaries of creativity.

The Role of Technology Companies

Companies may be held responsible for misuse of AI for generating harmful or illegal content and respecting copyright laws.

Companies need to be transparent about how their AI systems make decisions and the data they use to make those decisions. Many tech companies are now developing their own ethical guidelines for AI. These guidelines typically cover issues like fairness, accountability, transparency, and the impact of AI on society.

Looking Ahead

Future copyright law may need to expand its definitions to accommodate AI-generated works, potentially recognizing AI as a tool or medium rather than an author. Laws might evolve to clearly define who holds the copyright for AI-generated works.

A key challenge will be balancing the protection of creators’ rights with the encouragement of innovation and creativity through AI.

Dall-E 3 prevents you from asking for art in the style of a specific artist. Is that enough to protect the produced images from infringing copyright?

Establishing clear standards for the attribution of AI-generated works can help in maintaining transparency and fairness. International collaboration is essential for harmonizing copyright laws and standards.

Conclusion

It is a challenge to define authorship and ownership of AI-generated content and to adapt the existing legal frameworks to the nuances of AI.

There is an ethical concern of how AI will impact the perception of human-created works. Generative AI offers opportunities for innovation. It has the potential to redefine the creative process, open up new avenues for artistic expression, and even democratize content creation.

We must ensure that the rights of human creators are not overshadowed. This involves protecting their intellectual property, ensuring fair compensation, and recognizing their indispensable role in the creative process. We need to find a balance where AI is a tool for enhancing human creativity, not a replacement.

Creators need to be compensated for their work or we may dissuade future creators from ever creating.

References

- [The Verge Article](https://www.theverge.com/2023/11/4/23946353/generative-ai-copyright-training-data-openai-microsoft-google-meta-stabilityai)