Hash a key against the person's full name, date of birth, and social security number or equivalent in a public/private pair. This assumes a valid way of issuing a key to the correct person.
It would probably be better to chip IDs like with credit cards and incorporate a trusted protection module where no one has access to the embedded key. I haven't thought this through about the potential implications, but they are likely vast.
The TPM chip is how secure boot on computers works, and how Graphene OS is able to send secure OTA updates to devices.
The chip itself just handles encryption, where a secret hard coded key is set in a register that cannot be read by the accessible communications interfaces. Instead it creates additional keys that are hashed from the secret key. It can then verify if information was hashed by a key that was computed from the secret key. It is intended to resolve the issue of storing keys and the issue of the private key existing in plain text form at some point on a computer that may be compromised.
It would be absolutely essential that the distribution of these chips is randomized and anonymous. In theory, the secret key could be tracked from time of manufacture in the fab through to end user.
The insidious use case of TPM in personal computers is that software could be installed that then uses the TPM to make a new encrypted communication path to an external server. You would still see the DNS record with web 2.0, but with web 3.0 you implicitly trust in a way that means you may only see a connection to somewhere like google or amazon, but they could open up and forward that connection to any third party without you getting a DNS record of the connection. Under this circumstance, the TPM could be used without any record visible on your device. At least that is how I understand ECH and why the primary server connection has always remained visible in DNS records up until now.