I'll be honest. Facial recognition was not the use case I expected to encounter when researching OpenLedger's small language model ecosystem.
Then I thought about it for ten seconds and realized I should have expected it immediately. Wherever there's a decentralized data marketplace, someone is training a biometric model on it.
My concern isn't whether the technical architecture can support facial recognition SLMs. It can. My concern is what happens to the data used to train them. Whose faces. Collected how. Licensed by whom. Under what jurisdictional framework.
OpenLedger's attribution chain tells you where data came from. It doesn't tell you whether the person in the dataset consented to being in it.
Those are different questions. The second one matters more and gets asked less.