Vishing hit 11% of initial breach vectors in 2025 while email phishing fell to 6%. See the Mandiant data, the attack pattern, and defenses that work.
Voice phishing, or vishing, became the second most common way attackers broke into organizations in 2025. Mandiant’s M-Trends 2026 report put it at 11% of all initial infection vectors. Email phishing fell to 6% over the same period, down from 14% the year before. The reason is simple. A live caller adapts in real time, and no email filter ever sees the attack.
The shift is sharp, and it comes from incident response work rather than survey responses.
| Initial infection vector | Share in 2025 |
|---|---|
| Exploited vulnerabilities | 32% |
| Voice phishing | 11% |
| Email phishing | 6% |
Exploits stayed in first place for the sixth year running. So vishing did not become the single biggest threat. What changed is its position next to email. For years, email phishing was the social engineering vector that mattered most. In 2025 it dropped to less than half the rate of voice.
In cloud setups the picture is starker still. Vishing was the single most common way in, at 23% of cloud compromises, ahead of stolen credentials and email phishing.
An email is a fixed object. It arrives, a filter inspects it, and either it reaches the inbox or it does not. A phone call is a conversation.
That difference matters more than it sounds. A caller hears hesitation and changes approach. They answer questions an email never could. They build rapport over several minutes, then make the request once trust exists. Social engineering has always worked this way. Voice simply gives the attacker more room.
There is also nothing to scan. No link, no attachment, no sender domain, no message body. Every automated control a company bought for email stops mattering the moment the attack arrives by phone.
Discover how Threatcop protects your workforce from modern cyber threats.
One campaign shows the method clearly. Google’s threat research team tracks the group as UNC6040, and their approach barely changed across dozens of victims.
The caller pretends to be IT support. They reach an employee, build a believable reason for urgency, then walk the person through approving a connected app. The employee believes they are fixing something. In reality, they just gave an attacker-controlled app access to company Salesforce data.
Two things stand out. First, nobody exploited a vulnerability. Google’s own analysis confirmed the attackers used Salesforce exactly as designed. They used a modified version of a real data tool, which the victim authorized. Second, the extortion often came months later, tracked separately as UNC6240. So the organization had no obvious signal linking the call to the demand.
Attackers also used these calls to bypass multi-factor authentication directly, by talking the target through an approval rather than stealing a code.
Many of these campaigns aim at IT support rather than ordinary staff, and the logic is uncomfortable. Helpdesk impersonation works the same way in chat, so the pattern is not unique to phone calls.
A help desk exists to help. Managers measure it on resolution speed. Its staff learn to accommodate people who sound stressed. An attacker claiming to be a locked-out leader on a deadline exploits the function working exactly as intended.
The common requests follow a pattern worth knowing. A password reset for someone who cannot reach their email. An MFA device re-enrollment for a “new phone.” Or removing a security control that is “blocking urgent work.”
Spotting a fake voice is not a realistic defense, especially now that voice cloning is cheap. So process beats detection here, and the usual countermeasures against social engineering still apply.
A program that only tests email measures a shrinking share of real risk. Vishing tests a different reflex. The pressure is live, and the social cost of refusing feels higher.
Threatcop Security Awareness Training (TSAT) runs simulations across several attack types, including voice. Staff meet a vishing attempt in a safe setting, before a real one arrives. Its Employee Vulnerability Score shows who acts on a call, not just who clicks a link, and those are often different people. Average breach time, meaning how long from first contact to a harmful action, shows how much room an attacker would really have.
Those simulations focus on the callback habit, which is the one behavior that works regardless of how convincing the voice sounds.
Email phishing has not disappeared. But the numbers show attackers moving to a channel where defenders have almost no automated coverage. The answer is not better call screening. It is a callback rule that does not depend on recognizing a voice. Apply it consistently at the help desk and in finance, and practice it often enough that pausing a call feels normal rather than rude. Organizations still testing only their email defenses are measuring the vector that fell, while the one that doubled goes unexamined.
Voice phishing, or vishing, is a social engineering attack carried out over a phone call. The attacker poses as someone trusted, such as IT support or a senior leader. They then push the target into sharing credentials, approving access, or changing a security setting.
In Mandiant’s 2025 incident response data, yes. Vishing accounted for 11% of initial infection vectors, while email phishing fell to 6%. Exploited vulnerabilities remained the single most common vector overall, at 32%.
Because no email is involved. Filters, authentication records, and link scanners all inspect messages. A phone call contains none of those parts. The only controls that apply are callback checks and the judgment of the person answering.
Attackers talk the target through approving it. Rather than stealing a code, the caller invents a believable reason for the prompt to appear. Then they ask the person to approve it, or walk them through re-enrolling a device the attacker controls.
A mandatory callback to a number from your own records, not one the caller supplies. Vishing awareness and simulation builds that habit through practice. It works even when the voice is cloned perfectly, because it moves the verification off the channel the attacker controls.
Adhish Chakma is a Senior Product Manager at Kratikal, where he leads product initiatives focused on cybersecurity and AI-powered solutions. With experience in product management and cybersecurity, he works on developing practical technologies that address evolving security challenges. His areas of interest include People Security Management, cybersecurity awareness, AI-driven security, email security, and human-layer risk. He is passionate about building security products that make organizations more resilient against emerging cyber threats.
Most email false positives are not detection failures. See what actually causes them, and how to cut them without...
Calendar invites bypass email filters and can auto-add without acceptance. See how the attack works and the settings that...
Automation handles volume, people handle judgment. See the difference between humans in, on, and out of the loop, and...
