That "Insurance Rep" on Video Might Be a Deepfake — and Your Medical File Is the Prize

Only about a third of people can spot a deepfake — even after someone warns them one's coming. When nobody warns them at all, most still get fooled. Those aren't numbers from some distant lab. Researchers found that when people were told to watch for a fake video, barely a third could catch it.

If someone called you on a crystal-clear video,

If someone called you on a crystal-clear video, said they were from your insurance company, and asked to confirm your medical history — what would stop you? For most of us, honestly, almost nothing. That's the bet fraudsters are placing right now. A recent piece from Programming Insider lays out how fake faces and fake voices are moving into healthcare — and your medical file is the target. So how did stealing a person's identity get this easy?

Start with the money. Deloitte projects that losses from impersonation fraud in the U.S. will more than triple in just a few years — from around twelve billion dollars to roughly forty billion by twenty twenty-seven. That's the size of the wave building.

And the fakes themselves are getting good. The article describes A.I. tools that can now generate synthetic X-rays, M.R.I. scans, and voice recordings that pass a first look. Large language models can write detailed clinical notes and patient histories in minutes. Those fake documents can back up a claim for treatment that never happened. For an insurance investigator, that rewrites the job. For the rest of us, it means the paperwork in your medical record could be describing a stranger.

Why can't the old fraud systems catch this

Now, why can't the old fraud systems catch this? Because they were built for a different crime. The article explains that legacy tools watch the network and the system — they look for double-billing or phantom charges. Deepfakes don't attack the network. They attack your perception — a familiar face, a trusted voice. And the security questions we've leaned on for years? The reporting says knowledge-based checks get bypassed more than half the time. Even one-time passwords — those codes texted to your phone — get beaten in about one out of four attacks.

There's a speed problem too. Fraudsters upgrade their tools in days. Hospitals and insurers buy new tools on budget cycles measured in months. One side sprints. The other side files a purchase order. The U.K. government predicted eight million deepfakes would be shared last year — up from about five hundred thousand two years earlier. That's not a trend. That's a flood.

Some people push back. They say detection is improving, and that faking a convincing M.R.I. still takes real skill. The article's answer is blunt. The tools that write fake therapy notes and diagnostic reports are free, everywhere, and need no medical training at all. The skill barrier didn't shrink. It collapsed.