Elias décrypte la recherche derrière les gros titres en langage clair.
Inspiré par cette histoire ? Transformez l’idée en votre propre création par IA en quelques secondes — gratuit pour commencer, sans carte bancaire.
Commencer gratuitementChoisissez un compagnon et découvrez son avis sur cette histoire
An AI model from OpenAI, Codex, has reportedly been used to jailbreak the yet-to-be-released Nintendo Switch 2, allowing a modder to demonstrate custom game content such as new playable characters in a modified version of Mario Kart World.
Sammy Hajhamid, a modder known as "pepsipu," stated that OpenAI's Codex played a significant role in developing the "ns2jb" jailbreak. Codex is an AI model built by OpenAI that translates natural language into code, making it a powerful tool for developers and, as this case suggests, for those exploring system vulnerabilities. For creators, this means AI tools are not just for generating images or text; they can also streamline complex coding tasks involved in reverse engineering or developing custom software.
Hajhamid explained that Codex assisted in the jailbreak process by helping to translate specific natural language instructions into the necessary code. This capability allows a modder to describe a desired system behavior or exploit in plain English, and the AI generates potential code solutions, significantly accelerating development time. While the full extent of Codex's contribution to the "ns2jb" jailbreak remains under wraps, its reported involvement underscores how AI can augment the skills of even experienced developers and security researchers, potentially lowering the barrier to entry for complex technical projects.
This application of AI contrasts with more common uses in AI art, where models like DALL-E or Midjourney generate visuals from text prompts. Here, Codex is used for code generation, a function that directly impacts software development and system interaction. Creators interested in scripting or developing custom tools for their AI art workflows might find similar AI coding assistants valuable for automating repetitive tasks or tackling unfamiliar programming languages.
A modded version of Mario Kart World shows custom characters like Peter Griffin and Cinnamoroll racing.
Image : The Verge / The Verge AI
The demonstration involved the Switch 2 running a custom version of Mario Kart World featuring characters like Peter Griffin, Cinnamoroll, and Steve from Minecraft. This immediate application of the jailbreak for game modification illustrates the potential for AI to facilitate creative expression within existing digital frameworks, even those under strict control. For AI-art creators who often navigate intellectual property concerns, this development highlights a new frontier where AI tools can enable the creation of derivative works or introduce new content into established ecosystems. The ability to quickly generate code for such modifications could accelerate the pace of modding communities.
This incident also raises questions for platform holders regarding digital rights management (DRM) and system security. If AI tools can aid in rapidly discovering and exploiting vulnerabilities, companies like Nintendo may face an accelerated challenge in securing their hardware and software. The rapid pace of AI development means that methods of protection and circumvention are evolving quickly, creating a dynamic landscape for both creators who want to customize their experiences and companies aiming to protect their intellectual property. The use of AI in this context could lead to more sophisticated tools for both ethical security research and unauthorized access.
As AI models become more adept at code generation and problem-solving, their role in reverse engineering and exploit development is likely to grow. This trend could empower individual creators with advanced tools to customize their digital environments, but it also prompts further discussion about the ethical boundaries and legal implications of AI-assisted exploits. Developers leveraging AI in their own projects may find value in exploring how these tools can enhance efficiency, whether for creative coding or for understanding system architectures. Learning to generate images with AI, for instance, often involves understanding how to structure prompts effectively, a skill that shares conceptual parallels with instructing a code-generating AI.