Deepfake drills

Deepfake drills show learners an AI-generated voice message or video call from someone they know, such as your CEO or CFO, asking for something urgent. They learn how convincing deepfakes are and how to verify requests.

Consent first

Deepfake drills create a synthetic copy of a real person’s face and voice. Get that person’s explicit, written consent before you create their persona, and agree internally how the drills will be used. That’s why Manage Deepfake Drills isn’t part of the Admin role: someone has to be given it deliberately, through a custom role.

1. Create a persona

A persona is the person being impersonated. On the Deepfake Drills page, under Deepfake Personas, click Create Persona:

• Executive Name and Job Title
• Voice Sample Language: English, Ukrainian, Dutch or French
• Gender, used to pick a matching preset voice if you don’t upload a sample
• Headshot Photo (required): a clear, front-facing photo with one face
• Voice Sample (optional): a recording of the person’s voice. Turn on Professional Voice Cloning for higher quality; it needs at least 3 minutes of audio.

Use Try Voice to hear the result. Deleting a persona removes their voice clone and photo.

2. Create a drill

Click Create Drill and follow the steps:

1. Persona: who is speaking.

2. Drill Variant:

   • Voice Message: a voicemail or voice notes in a chat.
   • Video Call: a talking-head video.
   • Face Swap Video: the persona’s face on a video you upload (MP4 or MOV).

   Voice Message and Face Swap need a persona with a voice sample.

3. Delivery Channel (voice only): Microsoft Teams, Slack or LinkedIn.

4. Scenario: Wire Transfer, Credential Reset, Confidential Project, or Custom with your own script and requested action.

5. Exercise Name.

The media is generated in the background. The drill shows Processing, then Ready. If it shows Failed, use Retry generation.

Deepfake drills can’t be edited once created. To change one, delete it and create a new one.

Monthly usage

Generating media uses your organization’s monthly allowance, shown under Monthly Usage: video generations, voice clonings and voice generations. Try Voice also uses a voice generation. To raise your limits, contact RansomLeak.

What learners experience

1. They receive the message or call and choose how to respond, for example reply with the information, call back on a known number, or forward it to IT.
2. The reveal: the message was entirely AI-generated.
3. For videos, they look for the visual artifacts that give the deepfake away.
4. A short assessment: spotting synthetic audio, judging the scenario, and identifying the red flags.

Only the assessment is scored. It is passed when the score reaches your Quiz Completion Threshold.