AI Deepfake BEC Drains €95M via WhatsApp

A real-world business email compromise-style fraud hit Italian bank Fideuram after criminals impersonated executives using WhatsApp and an AI-cloned voice on a follow-up call. Staff were persuaded to transfer funds to foreign accounts before the scam was discovered, showing how AI makes familiar impersonation tactics more convincing and costly.

Key findings

• Fraudsters stole €95 million from Fideuram by impersonating senior executives, starting with a WhatsApp message and escalating to a phone call using an AI-cloned voice.
• The requested action was a funds transfer to foreign accounts; the bank only discovered the fraud after transfers were made (with partial recovery reported).
• The incident illustrates how AI can amplify traditional impersonation scams by making voice and identity spoofing more believable.

Who’s being targeted

• Commonly targeted roles: Finance, Treasury, Accounts Payable, Payment operations, Executive assistants.
• Affected industries: Banking, Financial services.
• Attack channels: whatsapp, vishing.
• Impersonated: Intesa Sanpaolo CEO Carlo Messina and a senior lawyer.

Awareness takeaways

• Treat WhatsApp (and other consumer messengers) as high-risk for payment instructions; require verification via an approved internal channel before acting.
• Do not approve or execute high-value transfers based on voice alone; use call-back to known numbers and step-up verification for unusual requests.
• Add extra controls for “executive” requests that bypass normal workflows (urgency, secrecy, or new destination accounts), since these are common fraud patterns.

Red flags to watch for

• Request arrives via WhatsApp rather than normal corporate approval channels
• Pressure/urgency and secrecy around a large transfer
• Caller identity relies on voice alone (potential AI voice cloning)