When a Face Is No Longer Proof

Digital onboarding rests on two pieces of evidence: a document and a face. Generative AI can now produce both well enough to pass a one-time check, and at scale.

The same technology sits on the other side: AI reads the document, matches the face, follows the transactions and flags what does not fit. Fraud and defense are built from the same parts, and the advantage goes to whoever has more data and learns faster.

The keynote argues that verification has to become continuous: identity confirmed at the start, then throughout the relationship, through behavior, devices and transactions. That takes capacity on both sides: institutions able to run it, and supervisors able to judge whether it works and whom it excludes.

It ends on the trade-offs: friction against security, privacy against monitoring, and the customer wrongly refused. A check that turns away the right people is also a failure.