Switch 2 Cracked: Old Game Plus AI Turns Out to Be the Key

Rumors of a Nintendo Switch 2 jailbreak have finally seen substantive progress. According to modder Sammy Hajhamid (known online as "pepsipu"), he used an old game and OpenAI's Codex model to successfully run a custom version of Mario Kart World on the Switch 2, adding new characters such as Cinnamoroll, the red Angry Bird, Peter Griffin, and Steve from Minecraft. This jailbreak project, dubbed "ns2jb," demonstrates the potential of AI-assisted reverse engineering.

How the Jailbreak Was Achieved: An Old Game and AI Working Together

According to pepsipu, the core of this jailbreak lies in finding a vulnerability in an old game and using OpenAI Codex to generate part of the exploit code. As a code generation model, Codex can quickly produce usable assembly instructions or scripts based on natural language descriptions or existing code snippets, greatly lowering the barrier to reverse engineering. However, pepsipu did not disclose which old game was involved or the details of the vulnerability, so the claim currently remains a one-sided assertion.

Console Jailbreaking and the Exploit Chain

Console jailbreaking is essentially about finding an executable code path within a vendor-locked software and hardware environment that allows unofficial programs to gain higher privileges. Common entry points include system menus, browsers, and game save parsing. Attackers typically need to combine multiple vulnerabilities: one to trigger memory corruption, one to bypass address space layout randomization (ASLR) and code signing, and another to escalate privileges or achieve persistence. Such combinations are called "exploit chains." Old games are often chosen because their code has been reverse-engineered for years, making vulnerabilities easier to locate and the runtime environment relatively controllable.

OpenAI Codex and Reverse Engineering

Codex is a GPT-based code generation model capable of generating code in multiple languages, including assembly and scripts, based on natural language or code context. In reverse engineering, it can assist in quickly writing proof-of-concept exploit code, parsing binary structures, or generating ROP chain fragments, thereby shortening the time required to hand-write attack code. However, Codex itself does not have the ability to discover vulnerabilities; it relies on the targets and ideas provided by the user, and its output still requires manual verification and debugging.

Demonstration: Custom Mario Kart World

After the successful jailbreak, pepsipu showcased a custom-modded version of Mario Kart World. The video featured unofficial characters such as Cinnamoroll, the red Angry Bird, Peter Griffin, and Steve, indicating that the jailbreak can modify game memory or resource files. For the player community, this usually means more unofficial mods, cheat codes, and homebrew programs may appear in the future.

The Role of AI in Jailbreaking and the Controversy

The involvement of OpenAI Codex is the most closely watched aspect of this incident. Traditionally, console jailbreaking relies on security researchers manually analyzing firmware and vulnerabilities, which can take months or even years. AI can accelerate code writing and the construction of exploit chains. But this also raises controversy: could AI be misused to bypass copyright protection? Nintendo has historically taken strong legal action against jailbreaking, and this incident may prompt the company to strengthen its anti-jailbreak measures.

The Boundaries of AI-Assisted Exploitation

AI code generation models function more as an "accelerator" than a "discoverer" in vulnerability exploitation. They can generate candidate exploit code based on known vulnerability types and mitigations, but they cannot replace a deep understanding of the target firmware, memory layout, and hardware behavior. Moreover, the reliability of generated code is limited by the security research samples in the training data, and facing new protection mechanisms often requires extensive manual correction. Therefore, AI lowers the barrier to code writing, not the barrier to vulnerability discovery itself.

Credibility and Subsequent Impact

Currently, the claim comes solely from pepsipu's personal disclosure, with no independent third-party verification. Nintendo has not responded. If true, this would be one of the earliest jailbreak cases after the Switch 2's launch, potentially affecting the console's security reputation. For ordinary players, jailbreaking means more customization possibilities, but it also comes with account bans and legal risks. We will continue to monitor further verification progress.

One-line Summary

Modder claims to have jailbroken the Switch 2 using an old game vulnerability and OpenAI Codex.

Paragraph Summary

Modder Sammy Hajhamid (pepsipu) claims to have used a vulnerability in an old game and attack code generated by OpenAI Codex to successfully run a custom version of Mario Kart World on the Switch 2, under the project name ns2jb. The demonstration featured unofficial characters such as Cinnamoroll, Angry Bird, Peter Griffin, and Steve, indicating that the jailbreak can modify game memory or resources. The incident has sparked discussion about AI-assisted reverse engineering lowering the barrier to jailbreaking, as well as controversies over copyright and misuse. Currently, the claim is only a one-sided disclosure lacking independent verification, and Nintendo has not responded.