Your next customer may never have existed.

A synthetic identity is the Frankenstein of digital crime: a fabricated identity combining real and fictitious data — a valid tax ID with an altered name, new phone and email, a fake address and a patiently built history. Unlike classic identity theft, there is no victim to complain: no one gets the wrongful bill, no one disputes it. The synthetic identity opens accounts, builds credit and reputation for months — and then "busts out": drains everything, disappears, and the company discovers it lent to someone who never existed.

Why has this become THE threat? Because generative AI has industrialized identity fabrication. Estimates indicate that up to 80% of new-account fraud is already driven by synthetic identities, with global losses between US$ 20 and 40 billion per year. Sumsub recorded over 300% growth in synthetic-document fraud in a single year, and deepfakes jumped from 500,000 files in 2023 to a projected 8 million in 2025, already accounting for around 40% of biometric-fraud attempts. Brazil is at the epicenter: an 830% rise in deepfake cases in twelve months, concentrating nearly half of Latin America's facial-manipulation cases.

Detection works in layers that traditional KYC lacks. Identity consistency asks whether the tax ID, name, date of birth and history add up — or whether the identity has no past, with a phone three days old and an email that never appeared anywhere. The link graph reveals the production line: the same phone across 14 records, the same address for 40 "people", the device that already created 9 accounts. Document forensics compares fonts, microprinting and physical consistency against AI-generated documents. Liveness with injection and presentation detection faces the deepfake injected into the camera and the filmed screen. Behavioral biometrics catches robotic form-filling and the "candidate" reading off another screen. And maturation ML learns the synthetic lifecycle, from the dormancy that builds a limit to the typical pre bust-out behavior.

The summary is simple: the identity that never lived does not pass. Real life leaves years of depth — ties, history and temporal consistency — that the identity fabricated yesterday cannot simulate, no matter how perfect its documents. Kavuka's edge is the depth of Brazilian data to measure that depth like no one else, combined with a national link graph, Brazilian document forensics and deepfake defense. The result is the correct diagnosis: the synthetic blocked at the door, the factory dismantled at the source, the alert before the bust-out — and the loss finally with the right name, fraud treated as fraud and credit treated as credit.