Vishing attack
A vishing attack is a fraud operation executed over a voice call. The attacker impersonates a trusted party — a bank, an executive, a family member, IT support — and pressures the target into moving money, revealing credentials, or authorizing access. Modern vishing attacks increasingly open with an AI-cloned voice of a real, specific person.
A vishing attack is the execution of vishing — the discrete fraud operation that starts when your phone rings and ends when money moves or credentials leak. Security teams treat "vishing" as the category and a vishing attack as the event you have to detect, interrupt, and report. What makes the 2026 version dangerous is that the voice on the line is often an AI clone of someone you actually know.
The four stages of a vishing attack
Every documented vishing attack — from gift-card scams to the $25.6M Arup fraud — follows the same arc:
- Reconnaissance. The attacker gathers a voice sample of the person they'll impersonate. Thirty seconds from an earnings call, a conference talk, a podcast, or a TikTok is enough for a modern zero-shot voice clone. For institutional impersonation (a bank's fraud department), they instead study the institution's real call scripts.
- Contact and spoofing. Caller ID is spoofed to the real number of the impersonated party. The victim's phone displays "Mom," "CEO," or the bank's official line.
- Urgency framing. The script compresses time so the victim can't verify: an account "being drained right now," a wire "needed before the deadline," a relative "in an accident and needing bail." Urgency is the single most reliable tell — legitimate institutions almost never require irreversible action inside one phone call.
- Extraction. The ask is either money (wire transfer, gift cards, crypto) or access (one-time passcodes, password resets, remote-desktop sessions). Sophisticated attacks extract in stages across several calls to stay under approval thresholds.
Real vishing attacks on record
Our Deepfake Incident Database tracks vishing attacks with primary-source citations. Three that define the threat:
- Arup, Hong Kong (2024) — a finance employee joined a video call where the CFO and colleagues were all AI-generated, then executed 15 transfers totaling $25.6 million. The attack combined video deepfakes with cloned voices.
- The New Hampshire Biden robocall (2024) — an AI clone of President Biden's voice told Democratic primary voters to stay home. The FCC responded by declaring AI-voice robocalls illegal under the TCPA and fined the operator $6 million.
- Ferrari (2024) — an executive received calls from a convincing clone of CEO Benedetto Vigna. The attack failed because the target asked a verification question only the real CEO could answer — the cheapest defense on record.
The scale beyond headline cases: the FBI's IC3 attributed $12.5 billion in reported losses to internet crime in 2024, Sumsub measured a 245% year-over-year increase in deepfake-enabled fraud attempts, and Mastercard found 46% of businesses have already been targeted.
Vishing attack vs. related attacks
| Attack | Channel | Typical payload |
|---|---|---|
| Vishing attack | Voice call | Wire transfer, OTP codes, remote access |
| Phishing | Credential-harvest links, malware | |
| Smishing | SMS | Malicious links, "package fee" microfraud |
| Deepfake video call | Live video meeting | High-value transfers (Arup pattern) |
See vishing vs phishing for a deeper comparison of the two most-confused terms.
How to defend against vishing attacks
Layer three controls; each catches what the previous one misses:
- Behavioral: treat urgency as the alarm. Any unexpected call demanding immediate, irreversible action is a vishing attack until proven otherwise — no matter whose voice it is.
- Procedural: out-of-band callback. Hang up and call back on a number you already have. This single habit defeats caller-ID spoofing and every voice clone in existence. Organizations should require callback verification plus dual approval for any payment change initiated by phone.
- Technical: detection. For a recording — a voicemail, a "proof of life" clip, a suspicious meeting recording — run it through the free AI voice detector. It returns a synthetic-vs-real verdict with the generator family it likely came from. For consumer-side triage, the scam call screener maps voicemails against 20 known scam script categories. At enterprise scale, contact centers integrate the same detection into the live call flow — the pattern the banking playbook covers.
Frequently asked questions
What is a vishing attack? A fraud call in which the attacker impersonates a trusted person or institution to extract money, credentials, or access — increasingly opened with an AI-cloned voice of a specific real person.
What is an example of a vishing attack? The Arup case: $25.6M transferred after a video call with AI-cloned colleagues. The New Hampshire Biden robocall is the best-known political example.
How do you know if a call is a vishing attack? Unexpected urgency plus a request to move money, share codes, or bypass procedure. Verify with an out-of-band callback, and run any recording through the AI voice detector.
Are vishing attacks illegal? Yes — wire fraud in most jurisdictions, and the FCC ruled in February 2024 that AI-generated robocall voices violate the TCPA. See deepfake laws by country.
Related terms
- Vishing — the parent category
- Vishing examples — a catalog of real cases by script type
- Voice cloning — the enabling technology
- Voice phishing — the long-form synonym