SDSignal Desk

Someone hacked the Switch 2 by using an old game and AI

Oct 22, 2026, 10:21 AM · The Verge

Image: The Verge

A Switch 2 jailbreak built with help from OpenAI's Codex is a fun mod demo, and a preview of how AI shortens the road from curiosity to exploit.

Why it matters

The Nintendo Switch 2 appears to have been jailbroken. Sammy Hajhamid, known as pepsipu, says OpenAI's Codex helped build the "ns2jb" exploit. It's software-only and tethered, meaning it has to be launched from a game each time the console boots, and it relies on SmileBASIC 4, a title that lets players write and run their own programs, though another modder says that game is "not strictly needed."

A modder has already shown a custom Mario Kart World with characters like Cinnamoroll, Peter Griffin, the red Angry Bird and Steve from Minecraft. How big a deal it is remains unsettled, and people in the Switch 2 hacks community are still debating its real impact.

From the desk

Let's keep this in proportion. A tethered, complex jailbreak isn't something most players will run, and the immediate result is mostly delightful nonsense: Peter Griffin in a kart. The homebrew community has long been a source of preservation work and creative tinkering, and we don't think console modding is a moral panic.

The AI part is what makes this worth our airtime. A coding agent helping find and build an exploit for a locked-down, current-generation console is a signal about capability. Our read is that reverse-engineering work that has traditionally demanded deep specialist skill may now move faster with an assistant doing some of the grinding. That's genuinely useful for defenders and researchers too. The same tools that speed up a jailbreak can speed up a security audit.

But the downside scales just as easily. If AI meaningfully lowers the skill floor for exploitation, the stakes don't stop at game consoles. The same pattern applies to routers, cars, medical devices and anything else that ships with code nobody expected outsiders to read. Console makers also lean on locked platforms to fight piracy, and a faster exploit pipeline makes that harder.

The likelier near-term read is a string of follow-on exploits, which the reporting suggests is possible, especially if more modders use AI. I'm watching for Nintendo's patch response and for whether coding-agent makers say anything about how their tools handle exploit development requests.

Context

Tethered jailbreaks don't survive a reboot, which limits their practical reach compared with persistent exploits. Using a legitimate game with programming features as an entry point is a familiar tactic in console hacking. What's newer is a researcher publicly crediting an AI coding agent for helping build it.

Who feels it

Nintendo
A potential opening for more exploits on its newest console, and pressure to patch quickly.
Modders and homebrew developers
A foothold for custom software and creative mods, though the method is still complex.
Security researchers
Another data point that AI coding agents can materially speed up exploit work.
AI coding tool makers
Growing scrutiny of how their agents are used for offensive security tasks.

What to watch

  1. A Nintendo firmware update targeting the exploit
  2. Whether a persistent or simpler jailbreak follows
  3. Other researchers crediting AI agents for console or device exploits
  4. Whether SmileBASIC 4 sales or availability are affected

Read the original

Continue at the source.

The Verge

Companies: OpenAI