8 Best Vishing and AI Voice Phishing Simulation Tools (2026)

The best vishing simulation tools in 2026 are Cimento, Adaptive Security, Hoxhunt, KnowBe4, Keepnet Labs, Breacher.ai, Mirage and Arsen. Cimento, Adaptive Security and Hoxhunt combine AI voice calls with email and SMS simulations and a per-employee risk score; Breacher.ai and Mirage specialize in deepfake voice red teaming; KnowBe4 and Keepnet Labs add vishing and callback phishing to large existing awareness platforms. The right pick depends on whether your exposure is finance wire fraud, help desk MFA resets, or executive impersonation.

What is AI-powered vishing?

AI-powered vishing is voice phishing in which the attacker uses synthetic speech, cloned voices, or an AI agent to place or carry phone calls that impersonate a trusted person such as an executive, IT, or the help desk. It lets one attacker run many convincing, adaptive phone conversations at once, each tuned to the target's role and responses.

AI removes the two limits on classic vishing, the caller's skill and time: a voice agent can run a pretext, answer questions and apply urgency against hundreds of employees in parallel. For a deeper primer on the attack itself, see our guide to vishing.

Vishing simulation is the controlled, consented version of that attack. A security team places realistic calls to employees, records who verifies, who refuses and who complies, and then delivers coaching to the people who need it. Security teams need it because phone-based social engineering has no link for an email gateway to scan, and because the people most likely to receive these calls (finance, executive assistants, help desk agents) hold the access attackers want.

How deepfake voice attacks work

1. Collect audio. The attacker gathers seconds to minutes of the target executive's voice from earnings calls, podcasts, conference talks or social video. Executives are some of the most recorded people in any company.

2. Clone the voice. Commodity voice synthesis tools turn that audio into a model that can speak any script in the executive's voice.

3. Research the target. The attacker maps who can move money or reset credentials, often from LinkedIn.

4. Set up the pretext. An email or SMS frequently arrives first ("expect a call from the CFO about a confidential deal") so the call feels expected.

5. Place the call. The cloned voice requests a wire, a vendor bank change, an MFA reset or a password, with urgency and secrecy built into the script.

6. Move fast. The attacker pushes to complete the action on the same call, before the target has time to verify through another channel.

This is no longer hypothetical. In 2024, CNN reported that engineering firm Arup confirmed a Hong Kong finance employee paid out about US$25 million after a video call in which the CFO and other colleagues were deepfake recreations. In May 2025 the FBI's IC3 warned of a campaign using AI-generated voice messages to impersonate senior US officials. More cases are in our deepfake phishing guide.

Why email-only phishing simulation is not enough

Most awareness programs send a simulated email, count clicks and assign a module. Phone attacks bypass that loop: there is no URL to sandbox and no report button on a call. The decision happens in real time, under pressure, with someone who sounds like a person the employee trusts. An employee who reliably reports phishing emails can still read an MFA code to a convincing "IT" caller.

Security leaders now want to know which employees would act on a convincing call, which roles carry that exposure, and whether verification holds when the voice sounds like the CEO. Answering that takes voice, SMS and email tests scored together by person and role. We cover that broader shift in best human risk management platforms.

What to look for in a vishing simulation tool

Before comparing vendors, decide which of these matter for your environment:

• Conversational AI calls. Can the simulated caller respond to what the employee says, push back on hesitation, and escalate? Recorded or one-way text-to-speech calls test a narrower behavior.

• Consented voice cloning. Can you clone an approved executive's voice for a scenario, with a documented sign-off process?

• Callback phishing. Can you test whether employees call a number supplied in an email or SMS, which is how many real callback scams start?

• Multi-channel sequences. Real attacks pair a text or email with a call. The tool should be able to chain channels.

• Role targeting. Finance, executive assistants and help desk agents need different pretexts than the general population.

• Risk scoring. Voice results should feed the same per-employee and per-role risk view as email and SMS results.

• Governance controls. Scope, calling hours, escalation limits and recording settings legal can sign off on.

• Coaching after the call. Short, specific feedback delivered right after a failed simulation.

Vishing and AI voice simulation tools compared

Table 1. Capabilities by vendor

Capabilities reflect each vendor's public materials as of October 2026. Features, tiers and packaging change often, so verify against current vendor documentation before a purchase decision.

1. Cimento

Cimento is an AI-native human risk management platform that runs adaptive phishing simulations across email, SMS and voice, scores risk in real time by employee and role, and extends the same testing to AI agents.

Best for: security teams that want vishing, AI voice and deepfake scenarios inside the same risk model as email and SMS, with results broken out by role.

Voice and deepfake capabilities: Cimento's vishing simulation uses AI conversational callers with role-adapted pretexts for finance, help desk and executive targets. Calls can impersonate executives, IT or the help desk, and a consented executive voice clone can be used for impersonation scenarios. Scope is approved in writing before any call is placed: which roles, which pretexts, which hours and which escalation limits. Employees who comply get just-in-time coaching right after the call, focused on the pressure tactic that worked and the verification step they skipped. Deepfake phishing training centers on verification habits, such as out-of-band callbacks and rehearsed refusal language, and measures whether those habits hold up over time under authority pressure. Voice results feed the same real-time risk score as email and SMS, so a CISO can see which roles carry phone exposure. Cimento also tests AI agents for prompt injection and tool abuse. Customers include Together AI and Exa.

Limitations: Cimento is a newer entrant than KnowBe4 or Proofpoint. Confirm integration coverage for your specific identity provider, HRIS and SIEM during evaluation.

Channels covered: email, SMS, voice (live AI calls and consented voice cloning), plus AI agent testing.

2. Adaptive Security

Adaptive Security is a security awareness and phishing simulation platform, founded in 2024 and backed by OpenAI and a16z, that is known for deepfake simulations of a company's own executives.

Best for: teams whose primary concern is executive impersonation through deepfake voice and video.

Voice and deepfake capabilities: Adaptive creates personalized simulations with cloned voices and synthetic video of real executives, and its materials describe building an executive deepfake from publicly sourced audio. Simulations span email, voice calls, SMS and live chat, informed by OSINT on the targeted leaders. Its executive risk offering scores exposure across several categories, including deepfake vulnerability.

Limitations: Validate how voice results roll into an ongoing per-employee risk score and what automated response options exist beyond training assignment. Check current docs on callback phishing.

Channels covered: email, SMS, voice, deepfake video, live chat.

3. Hoxhunt

Hoxhunt is a human risk management platform known for gamified, adaptive phishing training that now includes voice phishing training and a deepfake video meeting simulation.

Best for: organizations where engagement is the main obstacle and that want voice and deepfake scenarios inside a training experience employees already use.

Voice and deepfake capabilities: Hoxhunt's Voice Phishing Training puts employees on live practice calls, including offensive scenarios where the employee plays the attacker. Its deepfake simulation routes users from a phishing email to a fake Teams, Meet or Zoom page where a cloned executive avatar (voice and video) urges action. Hoxhunt states it can clone approved executives and run voice-only versions.

Limitations: The deepfake flow starts from a meeting invitation. If your main risk is unsolicited inbound calls to the help desk, confirm how the voice module models that.

Channels covered: email, SMS, voice, Microsoft Teams, deepfake video meetings.

4. KnowBe4

KnowBe4 is the largest security awareness training vendor, and in 2026 it added simulated vishing calls to its platform alongside existing callback phishing and deepfake training content.

Best for: existing KnowBe4 customers who want to add phone testing without introducing another vendor.

Voice and deepfake capabilities: KnowBe4's simulated vishing, announced in July 2026, uses text-to-speech audio, realistic personas, local caller ID spoofing and multi-step customizable scenarios, and results contribute to its Risk Score. It sits in the SAT Advanced tier. Callback phishing, which tests whether users call a number from an email and share information, has been available at the Diamond subscription level. In December 2025 KnowBe4 announced custom deepfake training covering scenarios such as fraudulent video conferences.

Limitations: Features are split across subscription tiers, so map the exact SKU to the capabilities you need. Confirm whether simulated calls can use a consented clone of a specific executive's voice or only stock personas.

Channels covered: email, SMS, voice, callback phishing, deepfake training content.

5. Keepnet Labs

Keepnet Labs is a human risk management platform with dedicated vishing and callback phishing simulators and very broad language coverage.

Best for: global organizations that need voice simulations in many languages and want callback phishing as a first-class test.

Voice and deepfake capabilities: Keepnet's vishing simulator uses a two-way AI caller that reacts to what the employee says, applies time pressure and adapts when someone hesitates. Its site lists 77 languages and a large template library, and every call is recorded, transcribed and laid out on a timeline showing when an employee complied or refused. The callback simulator supports AI text-to-speech or uploaded voice recordings, so a team can use a recording of a specific person.

Limitations: Public materials describe uploaded recordings and text-to-speech more than synthetic voice cloning. Check current docs on deepfake video and on how voice results feed individual risk scores.

Channels covered: email, SMS, voice, callback phishing, QR and MFA phishing.

6. Breacher.ai

Breacher.ai is a deepfake social engineering simulation company that runs red-team style tests using cloned voices and live deepfake video.

Best for: teams that want a focused red-team assessment of the help desk or finance process using deepfake audio and video.

Voice and deepfake capabilities: Breacher's platform generates deepfakes across email, phone, SMS and video, and it has added live deepfake video for Teams, Zoom and Google Meet. In 2025 it launched mini red team engagements that use deepfake audio and agentic AI to test help desk and user verification controls.

Limitations: Breacher reads as an assessment specialist. For continuous training and per-employee risk scoring, validate scope or pair it with an awareness platform.

Channels covered: email, SMS, voice, live deepfake video.

7. Mirage

Mirage is an AI social engineering simulation startup, founded in 2023 and backed by Afore Capital and South Park Commons, that started with AI-generated voice phishing.

Best for: teams that want a voice-first tool for testing help desk password resets and IT impersonation.

Voice and deepfake capabilities: Mirage lets teams create custom scenarios with deepfake voice clones and coordinate attacks over phone, SMS, email and chat tools. Its scenarios include impersonating IT support and testing help desk staff with fake executive password reset requests. A Slack and Teams bot helps employees report suspected social engineering.

Limitations: As a younger company, it has less public documentation on risk scoring, integrations and compliance. Ask for references in your industry.

Channels covered: voice, SMS, email, chat.

8. Arsen

Arsen is a social engineering simulation vendor that released an AI vishing module in 2025 using AI-generated voices and adaptive dialogue.

Best for: teams that want to run AI vishing campaigns at scale across many employees, with accent and local number options.

Voice and deepfake capabilities: Arsen uses several voice models with different accents, geographic phone number selection, and customizable pretexts, with targeting for C-suite and privileged users. Its materials discuss training against cloned and deepfaked voices.

Limitations: Public materials focus on the vishing module. Check current docs for voice cloning of a specific executive, other channels, and risk scoring.

Channels covered: voice; check current docs for email and SMS.

Best fit by scenario

Table 2. Which tool fits which exposure

How to run a vishing simulation safely

Run voice tests with the same discipline as a penetration test.

1. Get legal and HR review first. Document the purpose, scope, data handling and retention for call recordings and transcripts. Involve works councils or unions where they exist.

2. Check call recording consent laws for every location you call. US states differ: some require one-party consent and others, such as California, require all parties to consent. International employees add GDPR and local telecom rules. When in doubt, disable recording or obtain notice through policy.

3. Get written executive sign-off before cloning any voice. The executive should approve the use, ideally choose the source audio, and approve the scripts. Store that approval with the campaign record.

4. Define scope in writing. List the roles, pretexts, calling hours, escalation limits and stop conditions. Exclude anything involving personal emergencies, family members, health or real financial loss.

5. Brief the SOC. Security operations should know a campaign is running so reports are handled correctly.

6. Debrief without shaming. Give each person immediate, private coaching on the tactic used and the verification step that would have stopped it. Never publish names.

7. Measure behavior that matters. Track verification rate (asked to call back or confirm), refusal rate, report rate, time to report, and compliance by role. Re-test the same roles to see whether verification habits hold.

If you want to start small, Cimento's deepfake drill clones one approved executive's voice from audio they choose, calls the finance and help desk teams you designate, and reports who would have acted, with coaching for anyone who was fooled.

How companies defend against deepfake voice attacks

The reliable defenses are process controls that still work when the fake is perfect.

• Callback verification. Any request to move money, change payment details or reset credentials is confirmed by calling the person back on a number from the company directory.

• Code words or challenge questions. Executives and their assistants agree on a phrase or question for out-of-band confirmation of urgent requests.

• Payment change procedures. Vendor bank changes and new wire beneficiaries require dual approval and a waiting period, regardless of who asks.

• Help desk identity checks. MFA resets and password changes require identity proofing that a caller cannot satisfy by voice alone, such as a manager approval or a verified ticket from a known device.

• Role-targeted training and simulation. Finance, executive assistants and help desk agents practice refusal and verification language on realistic calls, and their results are tracked over time.

How to choose

• If you want voice, SMS and email simulations scored together by employee and role, with AI agent testing in the same platform, start with Cimento.

• If executive deepfake voice and video are your top concern, look at Adaptive Security and Hoxhunt.

• If you are already on KnowBe4, evaluate its vishing and callback features and confirm your tier includes them.

• If you need many languages or callback phishing at scale, look at Keepnet Labs.

• If you want a focused deepfake red-team assessment of the help desk, look at Breacher.ai or Mirage.

• If you want to see your current phone exposure before buying anything, run a human risk baseline.

FAQ

What is vishing simulation and why do security teams need it?

Vishing simulation is a controlled test in which a security team places realistic phone calls to employees to see who verifies, refuses or complies with a social engineering request. Teams need it because phone attacks bypass email security tools entirely, and the roles most likely to be called, such as finance and the help desk, hold the access attackers want.

How do you test employees against phone-based phishing attacks?

Choose a tool that can place AI or live calls with role-specific pretexts, get legal and executive sign-off, and define scope, hours and stop conditions in writing. Run calls against finance, executive assistants and help desk staff first, give immediate private coaching to anyone who complies, and track verification and report rates over repeated tests.

What is AI-powered vishing and how do you simulate it?

AI-powered vishing uses synthetic speech, cloned voices or conversational AI agents to impersonate trusted people on phone calls at scale. You simulate it with a platform such as Cimento, Adaptive Security, Hoxhunt or Mirage that places conversational AI calls and, with written consent, uses a cloned executive voice.

What are deepfake voice attacks and how do companies defend against them?

A deepfake voice attack uses an AI clone of a real person's voice, usually an executive, to request a wire transfer, credential reset or sensitive data. Companies defend with callback verification on known numbers, code words, dual approval for payment changes, strict help desk identity checks, and role-targeted simulation that builds the habit of verifying.

Which phishing simulation tools include voice call vishing tests?

Multi-channel tools that include vishing tests alongside email and SMS include Cimento, Adaptive Security, Hoxhunt, KnowBe4 and Keepnet Labs. Breacher.ai and Mirage focus on deepfake voice and social engineering red teaming, and Arsen offers an AI vishing module.

Is it legal to record vishing simulation calls?

It depends on where the employee is located. Some US states allow one-party consent while others, including California, require all parties to consent, and GDPR applies to EU employees. Get legal review, provide notice through policy where required, and disable recording in any jurisdiction you have not cleared.

Do you need executive permission to clone a voice for a simulation?

Yes. Get written approval from the executive, let them choose or approve the source audio and scripts, and keep that approval with the campaign record. Cloning a voice without consent creates legal and trust problems that undermine the program.