Fraud is targeting identity itself

Criminals now use AI to build false identities and mimic voices and faces well enough to pass existing checks. Banks must ask not just if a transaction looks risky, but whether the person behind it exists at all.

Scam rates are accelerating fast

Financial scams hit 40% of consumers in 2026, up from 34% a year earlier, with AI cited as a major driver. In the UK, scam attempts jumped 62% across over 100 million accounts at nine financial institutions.

Unauthorized fraud now dominates

Unauthorized-party fraud, where someone other than the account holder initiates the transaction, now accounts for 71% of incidents and dollar losses, up from 48% a year earlier.

Voice checks have already failed

Voice authentication, selfies, videos and behavioral biometrics are all vulnerable to AI mimicry. The deeper problem is behavior that looks ordinary every time it repeats, which is often the clearest sign it isn't human.

“voice authentication has already failed, leaving banks that lean on it exposed”

Banks are shifting to continuous trust

Instead of re-checking identity at every login, issuers are testing a model where trust builds session by session, and only unusual behavior triggers a harder look. Tokenization swaps static card numbers for credentials that specify what a customer can use them for.

Different fraud needs different defenses

A stolen card and a convincing fake customer need different defenses. Issuers must tell them apart in real time to protect customers without slowing them down.