Sharon Brightwell, a retiree from Dover, Florida, withdrew $15,000 after a cloned voice mimicking her daughter April claimed a car accident and arrest. The fraud, executed via AI voice synthesis, exemplifies a crisis now categorized distinctly by the FBI. In its 2026 Internet Crime Report, the bureau logged over 22,000 AI-enabled complaints with losses exceeding $893 million; $352 million targeted victims aged 60+. Total U.S. cybercrime losses hit $20.9 billion, a 26% rise. INTERPOL estimates global fraud losses at $442 billion in 2025, with AI fraud 4.5 times more profitable than traditional methods.

The core technology requires only three seconds of audio — a voicemail, TikTok clip, or Instagram video — to produce a near-perfect synthetic voice. Consumer Reports found most leading voice-cloning platforms (Descript, Eleven Labs, Lovo, Play HT, Resemble AI, Speechify) rely on self-attestation checkboxes rather than technical consent verification. Eleven Labs employs post-hoc safeguards like classifiers and traceability, but these activate after fraud occurs.

Detection-based defences are failing. Hany Farid, the foremost deepfake forensics authority at UC Berkeley, exemplifies the asymmetry: even experts struggle to identify clones in real time. With agentic AI now autonomously executing entire fraud campaigns, the economics of deception have inverted — industrial-scale impersonation costs nearly nothing and yields fortunes.