A Multilingual Deepfake Opportunity

At the recent Microsoft Ignite 2024, one product announcement stood out whilst sounding innovative and productivity-enhancing, it could also be a harbinger for a new wave of multilingual voice-based fraud vectors.

Microsoft announced the new feature, Interpreter in Teams, a tool that will allow Teams users to let Interpreter simulate their voices in 9 languages, with plans to expand to 31 different languages in a real-time Speech-to-Speech translation capability.

So, for multi-national organisations with employees speaking a range of native languages, or for doing business with a potential client that speaks Japanese, for instance, this capability would definitely be an aid to effective communication. But, as with so many well-intentioned developments, there can also be unintended consequences.

Deepfake fraud, the use of generative AI to clone images, video, and/or audio of a target is already rife, with CEO fraud already targeting at least 400 firms a day and according to a business.com study, more than 10% of companies have dealt with either attempted or successful deepfake fraud.

As generative AI tools used to create deepfakes become not only more advanced but also more readily available, these relatively new fraud vectors are only going to increase in both sophistication and volume.

How Voice Phishing Can Exploit AI Multilingual Innovation

The Interpreter in Teams, whilst scarce on detail at this stage, has the theoretical potential to expand the horizons of deepfake fraudsters. It doesn’t require a lot of imagination to picture a fraudster with access to the highly available Teams to create audio in different languages for voice phishing purposes in other countries or even inputting deepfake audio of a target and translate it into that target’s native language.

Fraudsters are some of the most imaginative, creative, and determined people around and there’s no doubt a tool such as Interpreter will already have them thinking about exploiting it. As a McAfee study found, 70% of people surveyed said they aren’t confident in telling the difference between a real and cloned voice.

AI Detection Is Critical for Business Security

Leaving deepfake detection to the hearing and interpretation of employees is clearly not an effective solution. For organisations to properly protect themselves against deepfake fraud attacks requires AI deepfake detection on every audio communication channel, from chatbots, online conferencing, and collaboration tools through to the contact centre or internal phone network.

Voice Verity^® is a non-biometric AI solution based on large-scale Deep Neural Network (DNN) techniques. It requires no enrolment, no consent, and no PII. Operating in both real-time and as a background process, it can be integrated into any channel carrying audio and requires just a snippet of audio to analyze and respond with a probability score of that audio being human or synthetic, regardless of language.

Only AI can effectively detect AI generated voices and only Voice Verity^® can provide the most accurate, omnichannel deepfake detection solution available. Deepfakes are here to stay, and the threat is increasing daily. Organizations urgently require detection strategies. The business.com study also found that damages from successful deepfake attacks accounted for up to 10% of the affected companies’ annual profits.

Reliance on human detection or a single communication channel is not a realistic defence. Protect every call, every chat, and every conference with Voice Verity® and fight AI with AI.