I assume, per the shared responsibility model, this implies guest OS (your instance) requires you to patch. Underlying host, hypervisor, would be patched.
But, perhaps that assumption was wrong and/or Amazon has another workaround.
If the hypervisor has the patch, but the Guest OS does not, it means that when the hypervisor context-switches into the Guest OS, it'll move its own kernel memory out of range of the exploit. So far so good. But the Guest OS is un-patched; when its kernel context switches into a userland process, its own kernel memory space remains available for the exploit. Thus, a patched hypervisor can protect itself, but not its guests.