DetectifAI builds on-phone tool to spot deepfake voices during calls

In short: DetectifAI is building voice deepfake detection that runs directly on smartphones, and the startup will compete at TechCrunch Disrupt.

What happened

DetectifAI is a San Francisco startup founded by Tarini Padmanabhuni after her grandfather was scammed by a fake voice. He received a call that sounded like his brother and was told to pay a ransom. Later, the family learned the voice was a “deepfake,” meaning an AI-made copy of a real person’s voice.

Padmanabhuni says the problem is growing quickly. TechCrunch cited FBI figures saying Americans lost close to $900 million to AI-driven scams last year, up 24% from 2024. Older adults were hit hardest, with people 60 and older losing about twice as much as those aged 50 to 59.

Many voice detection tools today run “in the cloud,” meaning they send audio to remote computers owned by a company, then send back a result. DetectifAI says it is taking a different approach by building smaller AI models that can run on the phone itself, like a spell checker that works even when you have no internet.

The company plans to sell first to phone manufacturers. It offers an SDK, which is a pack of ready-to-use code that phone makers can plug into their software. Padmanabhuni says DetectifAI already handles more than 100,000 calls a month for financial institutions in India, and it also checks “speaker verification” (confirming a caller is who they say they are).

DetectifAI has raised a small seed round and was selected to compete in TechCrunch’s Startup Battlefield at TechCrunch Disrupt, held October 13 to 15 in San Francisco.

Why it matters

If deepfake detection is built into phones, people may get warnings while a scam is happening, instead of finding out later. That could be especially helpful for families targeted by urgent, emotional calls.

Source: TechCrunch AI