How a Grandfather's Deepfake Scam Inspired DetectifAI's Real-Time Voice Fraud Detector
TL;DR
- Founder Tarini Padmanabhuni started DetectifAI after her grandfather was tricked by an AI-cloned voice of his brother in an emergency scam call.
- The San Francisco startup is building lightweight detection models that run on-device on smartphones to flag AI-generated voices in real time during calls.
- DetectifAI was selected as a Startup Battlefield competitor for TechCrunch Disrupt, putting its real-time voice fraud detector in the national spotlight.
A Scam Call That Hit Home
It started with a panicked phone call. Tarini Padmanabhuni's grandfather picked up to hear what sounded exactly like his brother — same cadence, same accent, same familiar warmth — pleading for urgent financial help.
It wasn't him. It was an AI-cloned deepfake of his voice, and by the time the family realized what had happened, money had already been sent.
For Padmanabhuni, then a young AI builder already fascinated by speech models, the incident was transformative. She had seen deepfakes as a technical problem. Now it was a deeply personal one, targeting exactly the kind of person least equipped to spot it: an elderly loved one who trusted what he heard.
That experience became the origin story for DetectifAI, the San Francisco-based startup she founded to fight AI voice fraud at the moment it happens.
The Scale of the AI Voice Fraud Crisis
DetectifAI is launching into a rapidly escalating crisis. AI voice-cloning tools can now replicate a person from just a few seconds of audio pulled from social media, voicemail, or a YouTube video. The FBI, FTC, and AARP have all warned of a surge in grandparent scams, kidnapping scams, and CEO fraud powered by synthetic voices.
Traditional defenses — carrier spam labels, post-call warnings, or cloud-based analysis — are often too slow or too invasive for live phone scams, where victims may have only minutes to decide. Padmanabhuni's bet is that detection has to happen in real time, on the call itself, before the damage is done.
How DetectifAI Works On Your Phone
Unlike heavyweight enterprise security systems, DetectifAI is designed to be lightweight and consumer-first.
The startup is building small, efficient AI models that run directly on smartphones rather than sending sensitive call audio to the cloud. During a live call, the system analyzes acoustic fingerprints and artifacts invisible to the human ear — unnatural breath patterns, spectral inconsistencies, and generative glitches left behind by text-to-speech and voice-conversion engines.
If the voice shows signs of being synthetic, the user gets an instant, on-screen alert flagging the call as potentially fake.
The on-device approach serves two purposes: speed and privacy. By processing locally, DetectifAI can warn users in seconds without latency, while keeping private conversations off remote servers. The company says its models are optimized to work on everyday iOS and Android devices without draining battery or requiring special hardware.
A Mission Focused on Protecting the Vulnerable
Padmanabhuni has framed DetectifAI's mission explicitly around protecting families, seniors, and first-generation immigrants — groups disproportionately hit by voice-based social engineering.
In interviews, she has described watching her own family navigate shame and confusion after the scam, a common aftermath that often goes unreported. DetectifAI, she says, isn't just about detection accuracy, it's about restoring trust in a phone call.
The startup is initially focused on consumer protection for personal calls, with longer-term plans to license its real-time detection SDK to banks, fintech apps, telecom carriers, and call-center platforms drowning in AI-powered fraud.
Spotlight at TechCrunch Disrupt
That mission and technical approach earned DetectifAI a coveted spot as a Startup Battlefield competitor at TechCrunch Disrupt in San Francisco.
Being selected for Startup Battlefield puts the young company on one of tech's biggest launch stages, pitching live in front of VCs, operators, and media alongside 200 other early-stage startups. For a company tackling consumer trust in AI, the timing couldn't be better, as voice fraud has become a headline topic for regulators and telecoms alike.
The team will be demoing its live detection on stage, showing how the app can distinguish a real human voice from a clone in real time.
What's Next for DetectifAI
Like many early-stage AI safety startups, DetectifAI now faces the classic challenges: staying ahead of rapidly improving voice generators, expanding across languages and accents, and keeping false positives low enough that users don't ignore warnings.
Deepfake generators evolve weekly, which means detection models must be constantly retrained. Padmanabhuni says DetectifAI is preparing for that arms race by training on a diverse dataset of commercial and open-source cloning engines, as well as real-world noisy phone audio rather than just clean lab samples.
If it succeeds, the granddaughter inspired by her grandfather's scam call could help define a new layer of defense for the AI era — one that lives quietly in your pocket, listening for the fakes you can't hear.
Get All The Latest Updates Delivered Straight To Your Inbox For Free!