Hardware attestation for end-to-end encrypted AI inference
Hype? Is it secure? Does it make sense for some use-cases, or is local AI strictly better?
Contains results of my audit, also mentioning @s...
i haven't looked at this yet, but it's very difficult to do correctly. TEE's are good for second party trust (eg the service runner -> hardware), but weak for third parties (eg, i can trust that venice can't see my data)