Deepfakes are a new and upcoming threat to business execs everywhere. If you're not familiar with the term, deepfakes are a type of artificial intelligence-generated images and videos that are designed to look and sound eerily realistic. This means that the person you're talking to on your next Zoom call might be an AI-generated deepfake.

This could have major implications for business execs, who might be tricked into revealing sensitive information or making deals with fake companies.

The technology could be used to create realistic and convincing fake video calls, which could be used for a variety of nefarious purposes.

For example, scammers could use deep fake live video calls to trick people into giving up their personal information or money. Furthermore, political enemies could also use this technology to spread misinformation or sow discord.

How does it work?

Deepfakes are made by feeding AI algorithms a large amount of data, such as images or videos of a person or celebrity. Then, the AI learns to generate new images or videos that look realistic enough to fool people into thinking they are speaking to a real person.

Here are a few easy ways to tell if you're talking to a deepfake:

  • Check for odd facial expressions or movements. Deepfakes often have difficulty replicating human facial expressions accurately.
  • Listen for any strange changes in the person's voice. Does it sound robotic or unnatural in any way?
  • Pay attention to the background. Is it too perfect or does it look like it's been edited?
  • Look for inconsistencies in the person's appearance, such as their hairstyle or color changing from one frame to the next.
  • Check if the person's behavior is erratic or if they seem to be avoiding eye contact.

How to check if you are talking to a deepfake?

As the potential for deepfakes to scam people during live video calls continues to rise, one company has found that AI-powered technology has a flaw that makes it easy to spot a fake.

Metaphysic, an AI content-generating company, recently put real-time deepfakes to the test and found that the technology can render a celebrity's face over a fake person's face during a video call only when they are facing forward. This means that if the person turns their head to the side or bends down, the deepfake will not be able to seamlessly transition and may look distorted as a result.

The scam is easily detectable when the user turns their face at a 90-degree angle or places a hand over their face. However, the company did note that the technology is still in its early stages, and faces rendered at sideway angles may improve over time as the software is refined.