How AI has been helping criminals who use deepfakes and voice cloning for financial scams, forcing banks and fintechs to invest in AI to counter fraud
Siddharth Venkataramakrishnan / Financial Times : X: @svr13 . LinkedIn: Anne-Marie Edmonds X: Siddharth Venkataramakrishnan / @svr13 : Amid all the guff of AGI, the experience of Paddric - who was targeted by scammers last year using a voice clone of his daughter - was one of the most disturbing stories I've reported on recently. AI heralds the next generation of financial scams https://www.ft.com/... [image] LinkedIn: Anne-Marie Edmonds : Great to see Alex West, our banking and payments fraud mitigation specialist providing his insight in the Financial Times. …
Context & Ripple Effects
AI-enabled impersonation in financial fraud has moved from an isolated corporate-loss case—a reported cloned-executive-voice theft—to tests of consumer-facing bank controls. Related coverage found that an AI replica could bypass a bank's voice-verification process, making the risk operational rather than merely theoretical.
This report matters because it identifies the corresponding defensive response: banks and fintechs are putting AI into fraud detection as synthetic voices and deepfakes make familiar authentication signals less reliable.
First-order effects
- Banks and fintechs must strengthen fraud controls against deepfake- and voice-clone-enabled scams, while customers face more scrutiny around high-risk instructions and identity checks.
- Criminals gain a more persuasive impersonation tool, particularly where victims or systems treat a familiar voice as proof of identity.
Second-order effects
- Voice-biometric and other single-signal authentication systems face pressure to add layered checks, after prior reporting showed an AI clone could defeat a bank voice-biometric system.
- Fraud-prevention vendors and financial institutions will compete on detection accuracy while balancing the added customer friction that stronger verification can create.
Third-order effects
- If synthetic impersonation keeps improving, financial identity is likely to shift away from voice or video as standalone evidence toward continuously assessed, multi-signal risk decisions.
- The pattern is part of an adversarial AI cycle: wider access to generative tools expands both fraud capability and demand for AI-based defenses, though the relative effectiveness of each side remains unsettled.
The trend: Financial services are moving toward agent-aware payment risk as AI makes human likeness a less dependable authentication signal.