The Race to Develop Voice AI Faster Than Scammers Can Clone It  

Generative artificial intelligence (AI) plays three notes: text, voice and video. 

And with all the hype around text-based generative AI capabilities, it’s easy to forget about the other two. 

But voice AI, particularly voice biometrics and authentication, is increasingly coming into vogue. 

PayPal is reportedly seeking to patent a novel way of tapping AI-based methods of voice authentication within customer service call settings, while Meta is experimenting with ways to identify and authenticate users across its platforms using their voice as a biometric key. 

Connected assistants like Amazon’s Alexa, Apple’s Siri and Google’s intelligent home devices have been tapping voice AI for years, but new generative AI solutions have upped the ante for connected audio. 

That’s because voice-based engagements, particularly around authentication, are emerging as a key innovation in financial services and payments, as well as other sectors. 

Still, building a 12-foot wall using one cybersecurity innovation generally leads to a 13-foot ladder from attackers, and voice AI is becoming incredibly simple to synthesize. 

See alsoWhy Whack-a-Mole Risk Prevention Won’t Work in Today’s Data Economy

Voice AI Could Give Fraudsters More Tricks

Biometric confirmation by way of voice AI has the potential to be a first-order part of a multifactor authentication system. Its hands-free ease of use could help financial services providers and FinTechs meet a new generation of end-user expectations around convenience, speed, accessibility and personalization. 

Password-free security is on the rise as organizations look to confirm the uniqueness of their users while establishing behavioral confidence bands through long-running interactions that leverage AI. 

Voice makes accessing account services and other features much simpler by providing frictionless capabilities that eliminate the need to click through online interfaces or even wait in person at physical locations. 

PYMNTS research in the report “How Consumers Want to Live in the Voice Economy” finds that over half of consumers (54%) would prefer voice technology in the future because it is faster than typing or using a touchscreen. 

Just last month, Mercedes and Microsoft teamed up to build out the functional use of in-car AI. Since June 16, the carmaker has been beta testing an integration of the ChatGPT function with its MBUX Voice Assistant in more than 900,000 vehicles.

Voice simply makes the process of multifactor authentication more convenient. 

But that same usability extends to bad actors and cyberscammers. 

See alsoGenerative AI Gives Scammers More Tools and Greater Reach

Integrating Voice Biometrics Responsibly

Privacy is not something to overlook, nor are technical avenues for misuse and abuse. 

Research in “Payments Security Amid Uncertainty: Fighting Fraud And Crime With Digital Innovation Playbook,” a PYMNTS collaboration with Citi, details how firms can pinpoint vulnerabilities and strengthen security to better position themselves. 

Audio files and video recordings are an important part of the data that organizations review as compliance monitoring requirements around voice conversations have become stricter in recent years.

That’s because generative voice AI is becoming worryingly good at cloning and generating voice outright — a highly effective way of duping both scam victims and potentially compromising biometric gateways. 

At the beginning of July, a group of U.S. senators called on federal regulators to protect consumers from what they saw as a rash of AI-powered scams.

“Voice cloning adds a new, threatening dimension to these scams, allowing fraudsters to generate voice clips to convincingly impersonate friends, family or potentially even financial advisers and bank employees,” the lawmakers said in a published letter. “Hearing trusted voices amplifies the risks.”

Echoing that sentiment, Doriel Abrahams, head of risk in the U.S. at Forter, told PYMNTS, “Generative AI now gives scammers an easy and effective way of building confidence with their targets. … It can be very challenging, particularly for non-digitally native generations, to discern what’s real from what’s fake in today’s AI-driven landscape.”

Still, PYMNTS research finds that nearly 6 in 10 consumers (58%) would use voice technology for the ability to complete tasks faster, easier and more efficiently. 

That’s why, as with the use and integration of any AI tool, secure data management protocols and bulletproof workflows are necessary for voice AI to be done responsibly.