Draft Proposal for an AI Content Law: A Framework for Creative Use and the Protection of Rights
The rapid advancement of generative artificial intelligence has disrupted the creative landscape, challenging pre-existing legal categories. Given this progress, a specific legal framework is urgently needed to balance the promotion of innovation with the protection of individual rights. This article proposes elements for a draft law that considers the unique aspects of generative AI.
1. "Good Use" in Generative AI:
Similar to the concept of "fair use" in traditional copyright, a principle of "good use" should apply to generative AI. This would allow the creation of transformative works, parodies, critiques, and other forms of artistic expression that utilize pre-existing material (including images) without explicit permission, provided the use is reasonable and does not significantly harm the value of the original material. In the context of the image of public figures, the creation of LoRAs that celebrate or evoke iconic images could fall under this "good use."
2. Public Domain and Public Figures:
Public figures, especially those in the entertainment industry, are inherently exposed to media visibility. Promotional photographs, event photos, and studio session images are often produced for public consumption. These images, intended for exposure, could be considered part of a "de facto" public domain for certain creative uses, particularly when the intention is not direct commercial exploitation without added artistic value.
3. Creation Based on Constructed Public Image:
The creation of AI content based on the public image that a person has constructed and that resides in the collective memory of the public should be permitted. This includes both the iconic physical image (features, memorable appearance) as well as the memorable roles and scenes for which the person is widely known and remembered. The creation of LoRAs that capture the visual "beautiful melody" of an actress, as well as the evocation of their most significant performances, would fall into this category.
4. The Principle of No Harm and the Burden of Proving Malice:
For claims of violation of the right of publicity by public figures in the context of generative AI, the burden of proving "actual malice or intent to harm" on the part of the creator should be required. This aligns with the protection of free speech and recognizes the diminished expectation of privacy of public figures. The creation of LoRAs that enhance the public image of a celebrity, without defamatory intent, would be protected under this principle.
5. Facial Recognition and Identity Traits:
In the realm of facial recognition applied to AI-generated images, the law should consider the ability to identify not only superficial resemblance but also the unique "identity traits" of a person. The absence of these distinctive features in a generated image could be a factor in determining the lack of a direct representation and, therefore, a lower likelihood of infringement.
Conclusion:
A legal framework for generative AI must be specific and balanced. By considering "good use," the status of public images, the permission for creation based on the constructed public image, the need to prove intent to harm, and the identification of identity traits, creative innovation can be fostered while offering certain protections. This proposal seeks to initiate a discussion on the essential elements of such legislation.