When Tarini Padmanabhuni’s grandfather got a call, he thought he was speaking with his brother. The voice on the other end claimed that his brother had been kidnapped and demanded a ransom for his release. The surprising detail? Her grandfather paid, completely unaware that his brother was safe and sound elsewhere. That voice was a deepfake–an AI-generated imitation of his brother’s speech.
Looking back on the incident, Padmanabhuni stated, “What stayed with me wasn’t the money. It was that he had no way of telling.” This experience motivated her to tackle the increasing threat of AI-driven scams, which the FBI reported cost Americans nearly $900 million last year–a notable 24% rise from 2024. Individuals aged 60 and older lost twice as much as those in the 50-to-59 age group, indicating a particularly at-risk demographic.
To combat this issue, she established DetectifAI, a company based in San Francisco focused on preventing similar scams. The deepfake voice detection market is competitive, with players like Reality Defender, Pindrop, Resemble AI, Microsoft Azure AI Content Safety, and Nuance. However, Padmanabhuni critiques current products, emphasizing that they typically rely on remote servers, which prevents phone manufacturers from integrating them directly into their devices. This situation leaves potential victims without sufficient defenses against such deceptive tactics.
Instead of creating large cloud models that must be scaled down for smartphones, DetectifAI develops compact AI models designed to operate within a smartphone’s operating system from the outset. This method enables real-time analysis of audio calls, voice messages, and other audio formats to determine if a voice is AI-generated, all without sending data to external servers.
DetectifAI plans to first engage phone manufacturers, licensing its software tools to ensure deepfake detection features are built into phone operating systems. Padmanabhuni compares this strategy to AT&T’s exclusive agreement during the original iPhone launch, arguing that the first phone maker to adopt DetectifAI could gain a competitive advantage. She believes deepfake detection will become as standard as camera resolution.
The company’s main product is its software development kit (SDK), which other companies can embed into their products. Beyond licensing to phone manufacturers, Padmanabhuni sees a secondary revenue stream through collaborations with businesses and fraud-prevention organizations.
Currently, DetectifAI is already generating revenue, reportedly managing over 100,000 calls each month for financial institutions in India. These calls are made by AI voice agents tasked with debt collection and loan document follow-ups. Each call employs deepfake detection and speaker verification to ensure authenticity.
Padmanabhuni began her machine learning journey at age 12 and later studied cyber-physical systems at Manipal Institute of Technology in India. There, she became the youngest team lead of India’s first driverless racecar division for Formula Student, an international engineering competition.
Reflecting on the most fulfilling moments for her startup, she highlights a small beta test conducted on WhatsApp. Users could forward suspicious voice notes and receive evaluations of their authenticity. One tester, whose relatives had been scammed, expressed a willingness to pay for the service without hesitation. “My grandfather didn’t have that,” she points out, underscoring the necessity for accessible verification tools.
So far, DetectifAI has secured a small seed investment from backers including Josh Constine, a former editor at TechCrunch, and Manohar Kamath, a principal at KM Growth. Additionally, the startup has been vetted by TechCrunch’s editorial team as a participant in the upcoming Startup Battlefield competition at TechCrunch Disrupt, scheduled for October 13 to 15 in downtown San Francisco.
The rise of deepfake technology poses a significant challenge, changing the way we communicate. As Padmanabhuni’s story shows, the ramifications of these advanced scams can be serious. Her work with DetectifAI could provide an essential layer of protection, equipping individuals with the tools needed to verify the authenticity of voices they encounter.
