FTC announces winners of Voice Cloning Challenge – and has a word or two for businesses

Voice cloning: A boon for people with medical issues or an AI-enabled bonanza for scammers? The answer depends on who’s using the technology and how they’re using it. That’s why the FTC is taking a 360° approach in the battle against the misuse of voice cloning. One part of that comprehensive strategy is the FTC Voice Cloning Challenge announced in November – and we’re excited to reveal the winners.

Impersonator scams bilked consumers out of a reported $1.1 billion in 2023. The FTC’s Trade Regulation Rule on Impersonation of Government and Businesses took effect on April 1, 2024, and offers new tools to take on that form of fraud. We’re also accepting public comments until April 30th on a supplemental notice of proposed rulemaking that would prohibit the impersonation of individuals.

Here’s how voice cloning fits in the picture. You’re familiar with family emergency scams where a person calls claiming to be a close relative who needs money ASAP for bail or hospital expenses. And you’ve heard about fraudsters posing as a “boss” calling an employee and directing them to transfer funds. As damaging as those scams already are, imagine if the voice on the other end of the phone sounds identical to a real family member or supervisor. That’s just one danger posed by the misuse of voice cloning technology.

The FTC will continue the law enforcement fight, but researchers, entrepreneurs, and tech experts have a role to play in the agency’s approach. In November we announced the Voice Cloning Challenge aimed at fostering breakthrough ideas on preventing, monitoring, and evaluating malicious voice cloning. And now – drum roll please – we’re pleased to announce the three winners who will split a $35,000 prize:

• AI Detect: This submission from OmniSpeech is aimed at consumer and enterprise apps and devices and would use AI algorithms to differentiate between genuine and synthetic voice patterns.
• DeFake: Ning Zhang from Washington University has suggested using a form of watermarking. The proposal calls for adding carefully crafted perturbations to voice samples that people would not be able to hear, but that could throw off AI voice cloners so that the audio could not be accurately cloned.
• OriginStory: Submitted by OriginStory LLC, this technology aims to authenticate at the time of recording that it’s a human voice by using off-the-shelf sensors already integrated into many devices. 

Although not eligible for the cash prize because of the size of the company, the fourth winning entry is from Pindrop Security. Its Voice Cloning Detection technology detects voice clones and audio deepfakes in real time. The technology evaluates each incoming phone call or digital audio in two-second chunks and flags those that are potential deep fakes.

The FTC thanks the panel of judges who evaluated the submissions: Professor Arvind Narayanan of Princeton University, Assistant Professor Britt Paris of Rutgers University, and Beau Woods, CEO of Stratigos Security and a Cyber Safety Innovation Fellow with the Atlantic Council.

Combatting AI-enabled fraud is a battle the FTC is fighting on many fronts and all of us have a stake in protecting consumers. We’re grateful to the winners, entrants, and judges of the Voice Cloning Challenge and look forward to continued engagement with academics, technologists, members of the tech industry, and law enforcement and government partners.

Two final thoughts for businesses. First, scammers target companies, too, so you have an interest in supporting efforts to stop the misuse of voice cloning. Second, scammers rarely operate alone. To profit from their schemes, they rely on the support of businesses that offer the additional services they need – platforms, payments, etc. Heed the warning that lending a hand to voice-cloning crooks could land you in legal hot water.