How Savi Security Built an AI-Powered Defense Against Voice Cloning Scams

Patrick and Ryan Coughlin, each with deep roots in enterprise cybersecurity—Patrick at Cisco and Splunk, Ryan at Apple and Spotify—have founded a startup specifically built to combat a new generation of fraud. Savi Security just closed a $7 million seed round led by Acrew Capital and is shipping its consumer app this week across iOS and Android.

The founders' motivation came from a chilling personal incident. Two years ago, Patrick's mother received a phone call that appeared to come from his sister, showing her number on the caller ID. The voice on the call sounded exactly like her daughter, crying "Mom, they've got me," followed by a man demanding $1,200 or threatening harm. The scammer had spoofed the number, synthesized the voice, and even knew details about a local Walmart the sister frequented. Patrick's mother avoided disaster only by calling her daughter directly to confirm she was safe—the entire incident had been an AI-powered scam.

The Economics of AI-Enabled Fraud#

What struck Patrick most was the shift in criminal economics. Historically, such elaborate impersonation attacks required expensive tools and deep reconnaissance, making them viable only against high-value targets like Fortune 500 companies or government agencies. Defending consumers wasn't worth the investment.

Generative AI and large language models have inverted that calculus. Voice cloning now requires just three seconds of audio from a public social media post. Personal details surface everywhere—in vacation photos, sports commentary, and public profiles. The barrier to entry for running a scam has collapsed while sophistication has skyrocketed, precisely mirroring the tools once reserved for state-sponsored attacks.

The Federal Trade Commission reported last month that consumers lost $3.5 billion to imposter fraud in 2025—triple the 2020 figure. While older Americans remain disproportionately targeted, research shows younger generations face unusual risk: Gen Z reportedly encountered text-based scams more frequently than other age groups and fell for them roughly 25% of the time.

Building a Real-Time Detector#

The Coughlin brothers began testing their vision through a free diagnostic tool called Scam Wise, which lets anyone anonymously upload suspicious texts, emails, or images for analysis. The site required no sign-up and generated immediate feedback on fraud likelihood. In four months, it accumulated 50,000 submissions and now receives roughly 10,000 new ones weekly—a goldmine of real-world data for training Savi's detection model.