vmblog logo 2024 wht (updated)

Recent Deepfake Attacks Are Exposing New Gaps in Corporate Security

Share: 

deepfake attacks corp security

This past February, a cloned voice pushed Fideuram, the private banking arm of Italy’s largest bank, Intesa Sanpaolo, to send about €95 million to fraudsters. This wasn’t a lucky scam. It relied on deepfake technology that was convincing enough to fool a senior banker. 

Attacks like this now go well beyond impersonation scams. Deepfakes are opening new paths for malware delivery, financial fraud, and corporate network infiltration. Three recent deepfake attacks show where defenses fall short and what organizations can do about it.

Fideuram’s €95 Million Fraud Exposes Weaknesses in Financial Verification

On February 23, Paolo Molesini, then chairman of Fideuram, received a WhatsApp message from someone posing as Carlo Messina, CEO of Fideuram’s parent group, Intesa Sanpaolo. The request seemed to come from the very top of the group, and it asked for urgent help with an overseas transaction.

A follow-up call used an AI-cloned voice of a senior partner at a prominent law firm, who had no part in the fraud. Molesini then instructed the finance department to arrange the transfers. Most of the money went to accounts in China and Hong Kong.

The attackers exploited trust in familiar identities and instructions that looked legitimate. That let them bypass the scrutiny an unusual financial request should get. Part of the money was later recovered abroad, but seven months on, at least €36 million is still missing. 

The lesson is simple. Voice recognition is not enough proof of identity anymore. High-value transactions need independent verification, such as a call back on a number the company already holds. Approval workflows should have no exceptions, even when the request seems to come from the CEO. 

BlueNoroff Shows How Deepfakes Can Lead to Malware Infections

The risk from deepfakes doesn’t end with financial fraud. These impersonation scams can also lead to malware delivery, potentially giving criminals a foothold in corporate networks. That’s what happened to an employee at a cryptocurrency foundation in 2025, in an attack that researchers attributed to BlueNoroff, a North Korean state-backed hacking group.

Attackers contacted the employee on Telegram, posing as outside professionals who wanted a meeting. A Calendly invitation then sent the employee to a fake Zoom domain the attackers controlled. 

The employee joined a group call that included deepfakes of the company’s own senior leaders, along with outside contacts. When the employee’s microphone failed to work, the fake executives said a Zoom extension would fix it. That “extension” was actually a malicious AppleScript file that opened a real Zoom page to look legitimate while quietly pulling a second-stage payload from a spoofed domain. The employee ran it, and the compromise began. 

The result was a full macOS compromise. Researchers recovered eight distinct malicious binaries, including infostealers, keyloggers, and backdoors. The malware could record keystrokes, capture the screen, and collect cryptocurrency-related files. It also embedded itself as a LaunchDaemon so it could survive restarts. 

The lesson is that deepfakes are not only a vehicle for social engineering. They can also open the door to initial access. In this case, no one asked the employee to send money. They only asked for a software install.

There were warning signs that security awareness training could have taught the employee to catch. The meeting link led to a fake Zoom domain. A request to install software came up mid-call. The right move was to leave and check with IT through a separate channel.

A Deepfake Job Applicant Reveals a New Insider Threat

In January 2026, Evoke Security CEO Jason Rebholz posted an opening for a security researcher position on LinkedIn. Within a couple of hours, a stranger messaged him to say he knew someone who would be a good fit. 

Rebholz scheduled a video interview, but it quickly became suspicious. The candidate’s face looked blurry and plastic-like, a green screen was reflected in his glasses, and his dimples appeared and disappeared. Rebholz recorded part of the interview and later confirmed with detection tools that the video was synthetic. 

The scheme matches a pattern security researchers have tied to North Korean IT workers. CrowdStrike found that the North Korea-linked group Famous Chollima infiltrated over 320 organizations in a single year, a 220% year-on-year increase. The group uses AI to generate resumes and run deepfake interviews. Hiring has become part of the attack surface. A fake hire gives criminal organizations legitimate corporate access, on top of a lucrative tech salary.

With the risk of deepfakes looming, organizations need stronger identity verification during recruitment and onboarding, especially for remote employees. Before granting anyone access, verify their government ID against the live person via a trusted process. Verify references independently, and treat unsolicited referrals with particular care.

Verification Is the Key to Stopping Deepfakes

You can’t train people to spot a perfect fake. You can train them to check. A callback on a known number, a second approver, or a quick message to IT takes a minute and works even when the fake is flawless.

Telling people to verify is not enough. They need to practice doing it under pressure. Run drills that look like the real thing, such as a CEO asking for an urgent transfer or a colleague on a video call asking to install something. Make deepfake scenarios a regular part of your security awareness program, and repeat them until they stick. Verifying must become a reflex, not an awkward extra step.