AI-generated voice and video impersonations fuel financial scams
Cybercriminals are using widely available AI models and online audio and video to impersonate trusted leaders and trick victims into handing over money. These scams can target employees with access to company financial systems, using fake calls or meetings to solicit credentials or transfers.

Key takeaways · 4
- 01
Voice-cloning systems can reproduce details of a speaker’s voice from only a few seconds of audio.
- 02
Attackers may use fake meetings to direct participants to links that provide access to company systems.
- 03
Deepfake attacks worldwide rose 165% year over year, with a peak of roughly 20,000 detected attacks in one month that summer.
- 04
If someone claiming to be a boss or executive calls, hang up and call back using a known number.
Impersonations are becoming more convincing
Criminals use online audio and video to create virtual copies of a CFO or another trusted leader.[1] Voice-cloning systems can reproduce a speaker’s timbre, cadence, accent, pauses and emotional tone from only a few seconds of audio.[1] Speech-to-speech AI can support real-time conversations with digital fakes.[1] Emerging video models can produce more consistent lip synchronization, gaze, facial movement, lighting and identity across frames.[1] The technology can also eliminate clues such as flickering faces, unnatural blinking and mismatched speech.[1]
The aim is access to money or systems
Impersonation schemes often target employees who can access company financial systems, seeking credentials or transfers to outside accounts.[1] AI can help fraudsters launch personalized, automated attacks at scale, faster and at lower cost.[1] Some generated impersonations can answer questions in real time and adapt to a conversation.[1] Downloadable tools can perform voice cloning and face swapping, among other tasks.[1] The North Korea-linked group described in the evidence was expanding beyond fake meetings to fraudulent document-sharing platforms, identity-verification workflows, media players and other business tools.[1]
Reported attacks include fake meetings
Deepfake attacks worldwide rose 165% year over year, reaching a peak of roughly 20,000 detected attacks in one month that summer.[1] Through August, the average was about one attack every three minutes.[1] A North Korea-linked hacking group used AI and publicly available photos and videos to create deepfakes of executives in cryptocurrency, Web 3.0, fintech and venture-capital firms.[1] During a meeting, participants could receive an error message and direct-message link that purported to help them rejoin but enabled access to company systems.[1] FBI investigators said the group had stolen more than $400 million since early 2025.[1]
Losses span multiple sectors
The FBI said hackers recorded meeting participants’ webcam images for future deepfakes and amassed more than 900 victim videos over the past year and a half.[1] Earlier that month, the Justice Department said it arrested three people accused of stealing more than $100 million from global companies using AI to impersonate corporate leaders.[1] The department said fake voice and video communications instructed midlevel employees to work with sham lawyers to transfer millions of dollars offshore.[1] A U.K. fintech firm wired $11.6 million, a Florida pet food company sent $5.2 million, and a Missouri hospital system sent $4.8 million.[1] The evidence warns that a familiar face or voice no longer proves identity.[1]
For teams handling payments, credentials or sensitive approvals, a convincing voice or video call is not enough to verify a request. A known-number callback offers a practical way to check an unexpected instruction before acting.
Why it matters
AI training for your team
kju.ai is built for teams: daily sessions, team analytics, and rollout support for your organization.
How this developed
21 November 2026
AI-generated voice and video impersonations fuel financial scams