No, you can reliably attest public source builds of critical software for the ultimate in transparency. That even includes models running on GPUs. Combine that with blind tokens and you get trusted, anonymous identity verification.
What you also get is mobile devices that can't run unblessed code, make it impossible to remove legally-mandated spyware or backdoors, as well as websites that you can't use anonymously, even when you have very valid reasons to do so.
They are implying the use of trusted computing with proprietary software to ensure that only users on fully “trusted” (locked down) devices are allowed to access network resources.
Presumably, if you have a trusted application on a trusted device, the identifier was installed in a trusted way, the device is in trusted possession and the device won't be given to anyone else, trusted computing may be able, in certain cases, to make it more difficult for a remote minor to use the identifier.