Artificial Intelligence, or AI, is an industry valued at nearly 300 billion dollars, according to Fortune Business Insight. Chatbots such as ChatGPT, Claude, and Gemini have become major players in digital life. Many companies, even those not related to technology, add the label "AI" to their products to increase their stock value. The phrase "AI-powered" is essential to establish a place in financial markets. AI is a highly controversial tool, with issues including copyright violations, the delegitimization of human creation, environmental impact, appropriation by political parties, or even use for military purposes. The technology has less than five years in the eyes of the general public and already has a complicated past. OpenAI, once at the top of the world, was overtaken by Anthropic and its agent Claude after a major controversy around a collaboration project with the US Department of Defense. The term "AI" encompasses a variety of different technologies, including robotics, facial recognition, and automatic doors. Some, such as smartphone facial recognition (Apple's Face ID implemented as early as 2017), can be considered "weak" AI with specific and limited tasks. However, when discussing "AI," especially "Generative AI" in the context of artistic fields, the focus is on two main technologies: the "LLM" (Large Language Model) and the "text-to-image" generators. LLMs are programmed to analyze data thousands or even millions of times to extract patterns and reproduce them. AIs like ChatGPT, Claude, or Gemini have absorbed millions of terabytes of information obtained from the web. Engineers gave them instructions to make them reproduce the rules of certain languages: word order in a sentence, association between a vocabulary word and its definition, grammatical structure. This set of instructions is called "Machine Learning." Text-to-image software such as Midjourney, DALL-E, or DeepAI mainly focus on the analysis of text in order to produce images. Unlike human illustrators, AIs do not follow a model of sketch, inking, and color. Instead, they produce results from what is called "random noise." This process leads to images that are initially blurry and gradually become clearer. Unlike illustrators, AIs generate results without caring about the concepts of anatomy, movement, or composition, which explains their tendency to create images with anatomical errors such as hands with six fingers or characters with three legs. To minimize the risk of error, it is more effective to indicate Hex codes rather than color names. An AI would prefer the instruction "t-shirt, clothing, short sleeves, color khaki HEX: #F0E68C" to "a short-sleeved yellow t-shirt." By the very nature of its functioning, which interprets the billions of images available in its database and constantly averages them, an AI always ends up producing results that resemble each other. This leads to the emergence of an "AI style." Today, any female character generated by AI is extremely sexualized, many characters have a slightly plastic manga style, and the networks are filled with videos of cute cats. Since the AIs have taken what was already abundant on the internet, they revolve around what is most present on the web: cats, anime, and pornography. This also leads to the reproduction and amplification of human biases. Journalists from Bloomberg have conducted tests by asking AIs to generate photorealistic images of human beings based on their professions. The finding was clear: a banker was mainly represented as a white man, a maintenance staff as a black woman. Therefore, AIs invisibilize entire parts of the population. Generating an image is far from neutral.