It's interesting.
Why "[wipe] the secrets used to decrypt the user's data, effectively destroying it" instead of wiping the data itself too?
Is this to potentially allow a third-party with enough power (i.e. a government entity) to eventually decrypt the data?
Destroying encryption keys stored in a small, secure processor is the most reliable and secure method.
The encrypted data can and should be wiped too, but it's a lot more reliable to have your security model not rely on that.
We have no reason to believe AES-256 will ever be broken within the lifetime of the universe. Maybe in 100 years I will be proven wrong, but I am willing to bet the security of my phone on it.
I suspect we will have human beings living in another solar system before AES-128 is broken.
Furthermore, as parent also noted, if your encryption is solid, wiping the keys is identical to wiping the data.
If you have a way to decrypt that data, you can just physically remove that memory, copy that data, and break the encryption on it. If people could break encryption, they wouldn’t go about updating the firmware and triggering a potential data wipe in the first place. In short, if you don’t trust encryption you’ve already lost.
It’s hard for me to an imagine a state funded attacker that wouldn’t just prevent the data wipe if they cared. I agree at least there’s an extreme edge case where this would protect you from a simultaneously very advanced but also negligent attacker, but that seems so rare that it’s unconcerning to me.
Feels a bit like adding an extra luggage lock to a bank vault since its still better to have more security afterall...
Or from the opposite direction: Only the keys are stored within the trusted part of the hardware; they're the only thing you can reliably wipe.