Nintendo Switch 2 Hacked with AI: Mario Kart World Modified

The Nintendo Switch 2 has been software-jailbroken for the first time. Security researcher Sammy Hajhamid used OpenAI Codex, among other tools, to uncover vulnerabilities and bypass security on Nintendo’s latest console. On Thursday, October 22, 2026, footage emerged of a modified version of Mario Kart World featuring characters such as Peter Griffin and Hatsune Miku behind the wheel.

The researcher, known online as pepsipu, has detailed his method on his website. His jailbreak, called ns2jb, reportedly also works on Nintendo Switch 2 firmware version 23.0.1.

The breakthrough opens the door to modifying games, running homebrew software and exploring the console more deeply. But the system’s security hasn’t been fully defeated: the jailbreak disappears after a reboot, and users still don’t have full control of the operating system.

For AI developers, the most intriguing part is how Hajhamid cracked the console’s defenses. He didn’t just ask AI to write code. He used multiple AI agents to hunt for bugs, run experiments and analyze the results.

The researcher used SmileBASIC 4, a program for the original Nintendo Switch that lets users create simple games. Thanks to backward compatibility, the software also runs on the Switch 2.

Codex uncovered a flaw in how SmileBASIC handles numbers and computer memory. Under certain calculations, a number can exceed the range the software expects—a problem known as an integer overflow.

Hajhamid used the vulnerability as a first step toward running code Nintendo normally blocks. He then found ways to access system components that manage files and load software, among other things.

The result is what’s known as a jailbreak. It lets developers run their own programs, back up saved game data and investigate how games and hardware work.

To show what he had achieved, the researcher modified Mario Kart World. His demo features new playable characters, including Peter Griffin from Family Guy, Hatsune Miku and Cinnamoroll.

Wccftech says it reviewed the footage and project source code and considers the jailbreak authentic.

Artificial intelligence did far more than write a few lines of code.

Hajhamid built an unusual test setup featuring both a Nintendo Switch and a Switch 2. He connected the consoles to equipment that let Codex view their screens and press buttons via USB.

He even added a remotely controlled button pusher. If an experiment caused a console to freeze, the device could automatically press its power button.

That setup let AI carry out some experiments on its own, investigate errors and prepare new tests. Hajhamid also had multiple AI agents analyze different parts of the security system.

Codex uncovered the SmileBASIC memory flaw and helped investigate other vulnerabilities. Hajhamid decided which avenues looked promising and worked on the final exploit himself.

It was no walk in the park. An earlier attempt to break in through the console’s browser failed to provide the access he needed. The researcher says the jailbreak took about two weeks of intensive work. This wasn’t a matter of simply giving ChatGPT a single prompt.

Notably, Hajhamid says Codex then programmed the modified version of Mario Kart World in a single day. The security research was complex; modifying the game itself was considerably quicker.

The discovery raises a compelling question: what happens as AI gets better at investigating and modifying existing software?

Security researchers typically have to comb through huge amounts of code to find bugs. Then they must determine whether those flaws can actually unlock features that are supposed to be protected.

AI could take on more and more of that work. Coding agents can analyze code, flag suspicious components, write test programs and adjust their approach based on the results.

The Switch 2 jailbreak shows how this can work in practice. But the same technology has potential beyond cybersecurity.

We previously covered a developer using AI to create new mods for Fallout 2. Players can describe the characters, items or quests they want to add, and AI helps build those changes.

Modern consoles are much harder to crack. Manufacturers like Nintendo protect their systems with multiple layers of security, keeping developers from freely accessing the files and features they might want to modify.

A jailbreak can change that.

Future AI agents could analyse an existing game, figure out how its different components work, and then suggest modifications of their own. That might mean new characters, extra gameplay mechanics or upgrades for older games.

These tools could also make it easier to reverse-engineer software. We’ve previously covered how AI is bringing old video games back to life.

That doesn’t mean AI will soon be able to crack every console automatically. New security layers, technical limitations and bugs in AI-generated code remain major obstacles.

Still, AI could speed up this kind of research. Work that once required specialised teams may increasingly be within reach of smaller groups.

For Nintendo, this development is less welcome.

The Switch 2 has only been available since June 2025. Nintendo is working to stop users from bypassing its security, running unauthorised software or playing pirated copies of games.

The new jailbreak doesn’t yet prove that any of those things are possible.

Users have to run the exploit again after every reboot. ns2jb also doesn’t provide full kernel-level access—the deepest level of the operating system. The discovery has not yet enabled a fully fledged alternative operating system for the Switch 2.

The method also uses SmileBASIC 4 as its entry point. Nintendo could try to close this route with security updates or by restricting access to the program.

The real risk for Nintendo lies in what happens next. Other researchers could build on the discovery and search for new vulnerabilities.

AI may play an increasingly significant role in that process.

This trend extends beyond gaming consoles. On Wednesday, AIWereld.nl reported on two critical PHP security flaws that AI helped uncover.

That makes the technology valuable to both security researchers and software companies. The same AI that can help exploit vulnerabilities can also help developers secure their products.

For players, the Switch 2 jailbreak mainly opens up new ways to experiment with games—for now. For Nintendo and other tech companies, it’s a warning that AI is increasingly being used to probe their security, too.

Perhaps the most striking development isn’t that a game console has been hacked. That has happened before. What’s different is that a researcher used AI agents to analyse code, run experiments and help find a working attack method.