Generative artificial intelligence, or generative AI, refers to a subset of artificial intelligence (“AI”) techniques and models that are designed to generate new content or data that is similar to existing examples. Unlike traditional AI systems that focus on recognizing patterns and making predictions, generative AI aims to create new content that has not been explicitly programmed or encountered before. Generative AI models leverage machine learning algorithms, particularly deep learning techniques such as neural networks, to generate new content based on patterns and structures learned from training data. These models, which have been built by the likes of OpenAI, Stability AI, and others, can learn the underlying characteristics and distribution of a dataset and then generate new samples that resemble the original data.
There are different types of generative AI models, each with its own approach and application …
Generative Adversarial Networks (“GANs”): GANs consist of two neural networks, a generator and a discriminator, that are trained in tandem. The generator creates new samples, such as images or text, while the discriminator tries to distinguish between the generated samples and real examples. Through an iterative process, the generator learns to create increasingly realistic content, while the discriminator becomes better at identifying generated content. GANs have been used for various applications, including image synthesis, video generation, and text generation.
Variational Autoencoders (“VAEs”): VAEs are generative models that aim to learn the underlying distribution of a dataset and generate new samples from it. They use an encoder network to map input data into a latent space representation, and a decoder network to reconstruct the input data from the latent space. VAEs can generate new samples by sampling points from the latent space and decoding them into the original data domain. They have been used for tasks such as image generation and data synthesis.
Autoregressive Models: Autoregressive models generate new content by modeling the conditional probability of each element in a sequence given the previous elements. These models generate content element by element, such as generating text one word at a time. They can capture complex dependencies and structures within the data. Examples of autoregressive models include language models like GPT (Generative Pre-trained Transformer).
Generative AI has found applications in various domains, including art, music, image synthesis, text generation, and content creation. And it is important to note that generative AI models have given rise to legal issues, including on the copyright and trademark infringement fronts, and can sometimes generate content that may be misleading, biased, or inappropriate. Ethical considerations and responsible deployment of generative AI are essential to mitigate potential risks and ensure the responsible use of these technologies.
As of 2025, the legal landscape surrounding copyright and generative AI remains complex and unsettled. On January 29, 2025, the U.S. Copyright Office released the second part of its Report addressing the copyrightability of AI-generated materials. It explores the nature and extent of human involvement required for these outputs to qualify for copyright protection under U.S. law. The Report delineates between “AI-assisted” and “AI-created” works, providing guidelines for what and what does not qualify for copyright. The Copyright Office emphasized that the copyrightability of AI-generated works must be determined on a case-by-case basis, considering the specific circumstances of their creation.
While AI assistance does not inherently affect copyrightability, works that are entirely generated by AI are not eligible for protection, failing the Feist standard because it lacks independent human creativity. Further, if an artwork is produced entirely by AI, inputting a prompt into an AI system does not provide enough human control to establish authorship as prompts merely function as instructions rather than expressions of creative authorship.
However, if an artist employs AI to generate elements of a piece but maintains creative control over the selection, arrangement, and modification of these elements, the work is considered “AI-assisted” and may be eligible for copyright. Further, when AI tools model a human-created, copyrightable work, the human author retains copyright over the unaltered, perceptible portions of their original authorship as well as the creative selection or creative modifications of the outputs.
Notifications