It was a ticking time bomb made of glass and silicon.
In 2018, early adopters paid $2,300 for the Magic Leap One, a futuristic augmented reality headset that promised to change the world. But by August 2023, the company had issued a death sentence: on December 31, 2024, the servers would go dark. Without them, a mandatory security check would fail, turning thousands of functional, high-end devices into expensive paperweights.
It was the ultimate example of planned obsolescence. Until a security researcher named Elise Amber Katze decided to intervene.
The Call to Action
The story starts in mid-2024, six months before the deadline. A friend approached Katze with a desperate plea: their expensive hardware was about to be bricked by a corporate kill switch.
For Katze, this wasn't just a technical challenge; it was a moral imperative.
"I morally dislike this," Katze told the audience at the 39th Chaos Communication Congress (39C3) in Hamburg. "Companies shouldn't turn functional devices into e-waste just because they want to sell newer devices."
With the clock ticking toward New Year's Eve, Katze went to work.
The Heist: Cracking the Silicon
The target was the NVIDIA Tegra X2, the system-on-chip (SoC) powering the headset. Katze didn't just want to bypass a login screen; she needed to own the hardware.
The attack unfolded in three cinematic stages:
- The Backdoor (Sparsehax): Katze discovered that Magic Leap’s software was built on NVIDIA’s open-source code. She found a flaw in how the system unpacked files, allowing her to smash the stack and inject code over a simple USB connection.
- The Foothold (Dtbhax): Getting in was one thing; staying in was another. She exploited the kernel’s loading process to ensure her jailbreak survived a reboot.
- The "God Mode" (BootROM): This was the nuclear option. Using fault injection—literally glitching the hardware with voltage—she dumped the BootROM. She found a vulnerability in the chip's read-only memory. Because this code is etched into the physical silicon, it is unpatchable. NVIDIA cannot fix it with an update.
The kill switch was defeated. The headset was saved. But Katze wasn't done.
The Plot Twist
In a classic documentary twist, it turned out the Magic Leap One wasn't the only device relying on the Tegra X2.
While digging through the code, Katze realized the same unpatchable vulnerability existed in another piece of hardware—one that moves at 70 miles per hour.
Tesla Autopilot.
During her presentation, Katze dropped the bombshell: the exact same exploit chain she used to save an AR headset works on Tesla’s Autopilot 2 and 2.5 hardware. She demonstrated that an attacker with physical access could bypass the secure boot on a Tesla, granting them total control over the Autopilot module.
The Aftermath
What began as a mission to stop a $2,300 gadget from becoming e-waste ended with the exposure of a fundamental flaw in one of the automotive industry's most critical chips.
The Magic Leap servers shut down on December 31, 2024, as planned. But thanks to Katze, the devices didn't die. They were liberated.
Top comments (2)
Is it the same vulnerability that caused Fusee Gelee in the v1 Nintendo Switch?
I guess not really, but they could be 'cousins.' Here is my breakdown; please feel free to share your thoughts below:
Different Silicon: Fusee Gelee exploits the Tegra X1 (Switch). This hack targets the Tegra X2 (Magic Leap One/Tesla).
Different Vector: Fusee Gelee attacked the USB recovery stack. This exploit targets the BPMP (Boot and Power Management Processor).
Same Outcome: Both are unpatchable BootROM exploits, meaning they compromise the hardware permanently and cannot be fixed via software updates.
Does anyone know if the Tegra X2 is widely used in other IoT devices besides Magic Leap and older Teslas? Curious how far this impact reaches.