DetectifAI launched the week €95M walked out of Fideuram on a cloned CEO voice — and Palo Alto Networks, CrowdStrike, Zscaler aren't selling one

On Feb 13, 2026, Fideuram chairman Paolo Molesini received a WhatsApp message appearing to come from Intesa Sanpaolo CEO Carlo Messina, asking for urgent help with an overseas transaction. A short time later, a phone call supposedly from a senior partner at a top law firm confirmed the request — the voice was an AI clone. Molesini instructed Fideuram's finance team to wire the money, mostly to accounts in China and Hong Kong. By the time irregularities surfaced, roughly €95M had already left the building.

Twenty-one-year-old Tarini Padmanabhuni founded DetectifAI after her grandfather was tricked by a deepfake of his brother's voice into paying a ransom. The San Francisco-based startup publicly launched on Sep 28, 2026, in a TechCrunch profile that also revealed early commercial traction: more than 100,000 calls per month running through Indian financial institutions for debt collection and loan-document follow-ups, with deepfake detection and speaker verification applied to every call. The founder declined to name customers, citing confidentiality.

The four names investors actually trade — Palo Alto Networks, CrowdStrike, Zscaler, and Cisco — have customer relationships with the same banks that just lost money to cloned voices. None of them ships a dedicated voice-channel deepfake detection product today. Palo Alto Networks's Unit 42 publishes technical analyses of deepfake scam operations, CrowdStrike's Charlotte AI agentic platform handles detection triage, and Zscaler's zero-trust architecture is the structural fit for voice-channel threats — but the actual \"is this voice real?\" decision is being sold by the pure-plays. Cisco is the only listed name with direct exposure, via its 2025 co-lead of GetReal Security's Series A.

Deloitte's Center for Financial Services put the number on the record in 2024: US losses from generative-AI-driven banking fraud are projected to climb from $12.3B in 2023 to roughly $40B by 2027 — a ~34% compound rate. The deepfake-AI detection market, narrower than the loss figure but the addressable spend for vendors, is sized by The Business Research Company at $1.29B in 2026 growing to $3.2B by 2030 (a 25.6% CAGR). A separate Market.us projection puts the same market at $5.6B by 2034 on a 47.6% CAGR — the wide range is itself the story: vendors and analysts don't yet agree on how fast this category compounds.

Short-term (next two quarters): the Fideuram attack will trigger procurement cycles at European private banks, where Palo Alto Networks, CrowdStrike, and Cisco all have named financial-services practices. The question is whether they answer with a bundled feature or an acquisition. Zscaler is the wildcard — architecturally the cleanest fit, but unprofitable on a TTM basis through Jul 2026.

Long-term (1–3 years): the structural shift is that voice-channel verification becomes a phone-OS feature, not a bank product. DetectifAI's SDK distribution model is built for exactly that — if it lands at one or two phone OEMs, every listed incumbent becomes a downstream customer rather than a competitor. The deeper question is who owns the \"trust\" layer for AI-generated audio: the OS vendor, the cybersecurity platform, the carrier, or the bank. The €95M Fideuram case is the first time the question has been priced into a public procurement decision.