When the phone rang, Tarini Padmanavni’s grandfather believed he was talking to his brother. The voice on the other end said he had been kidnapped and the only way to get him back was to pay the ransom. Her grandfather paid the money, but later learned that his brother was in a completely different place, completely unaware of what had happened. The voice turned out to be a deepfake, an AI-generated imitation.
“It wasn’t the money that stuck with me,” Padmanavni says of the incident. “He had no way of communicating that.”
That was about two years ago. Now, she says, the San Francisco-based company she founded, DetectifAI, aims to prevent others from being fooled in the same way.
This is a big, growing problem, and there’s a market for it. According to the FBI, Americans lost nearly $900 million to AI-powered scams last year, a 24% increase compared to 2024. People aged 60 and older suffered twice as much loss as those aged 50 to 59.
There is a lot of competition in deepfake audio detection from companies like Reality Defender, Pindrop, Resemble AI, Microsoft Azure AI Content Safety, and Nuance (also owned by Microsoft). But because today’s detection products run in the cloud on remote servers, cell phone manufacturers can’t build them directly into their devices, leaving those targeted with few defenses, Padmanavni said.
DetectifAI says it is designing compact AI models from scratch that are small enough to run within a smartphone’s operating system, rather than shrinking large cloud models to fit on a mobile phone as some companies do. The goal is to be able to instantly determine whether a call, voice message, or other audio is generated by AI, without the audio leaving the device.
DetectifAI will first sell to cell phone manufacturers, who will license the software tool so they can ship detection capabilities as a built-in feature of the cell phone’s operating system. Padmanabuni’s analogy is to AT&T’s role in the launch of the original iPhone, which set itself apart from its competitors through exclusive deals with carriers. The first phone manufacturer to ship DetectifAI will have an advantage over its competitors, she suggests.
At the core of the product is DetectifAI’s Software Development Kit (SDK). This is a package of code that other companies can incorporate into their products and license through existing channels. Padmanavni said a second source of revenue will come from licensing technology to corporations and anti-fraud companies.
Meanwhile, Padmanavni said the startup has already generated early revenue and is processing more than 100,000 calls a month from Indian financial institutions. These calls are made by AI voice agents that collect debts or follow up on loan documents, and every call goes through deepfake detection and speaker verification. Padmanavni declined to reveal the client’s name, citing non-disclosure agreements.
Padmanavni says she started working in machine learning at the age of 12, then studied cyber-physical systems (technologies that link software and physical machines) at India’s Manipal Institute of Technology, and became the youngest team leader for India’s first driverless race car division of Formula Student, an international student engineering competition.
DetectifAI has so far raised a small seed amount from investors Josh Constine (former TechCrunch editor) and Manohar Kamith, principal at consulting services firm KM Growth. The company is one of the startups vetted by TechCrunch’s editorial team to compete in the prestigious Startup Battlefield competition at TechCrunch Disrupt, which will be held in downtown San Francisco from October 13th to 15th.
To check out DetecifAI and the other startups participating in this year’s contest (and network with the people funding them), be sure to join us next month.
If you buy through links in our articles, we may earn a small commission. This does not affect editorial independence.
