💊 The $1.8B "AI Success Story" Built on 800 Fake Doctors

The New York Times called it the first billion-dollar company built by one person with AI. Then people started looking closer. Behind the headline were 800 fake doctors, deepfaked patients, an FDA warning, and a class action lawsuit. Let’s dive in!

In today’s insights:

LATEST DEVELOPMENTS

Evolving AI: A two-person telehealth startup hailed as the first AI-powered billion-dollar company is unraveling under fraud allegations.

On April 2 The New York Times profiled Matthew Gallagher as proof that one person with AI could build a billion-dollar company. He launched Medvi in late 2024 with $20K and AI tools to sell compounded GLP-1 weight loss drugs via telehealth. The company claims $401M in 2025 revenue with just two employees and projects $1.8B for 2026. But those numbers come from a private LLC with no audited financials and no independent verification. What has been verified paints a different picture. Futurism found that Medvi's patient before-and-after photos were deepfakes traced to old images with AI-swapped faces. Behind the growth were 800+ Facebook pages posing as real doctors to reach women aged 35-55. The FDA had already warned that Medvi's marketing falsely implied its drugs were FDA-approved and the company's own fine print admits people in its ads may be actors or AI and not licensed professionals.

Medvi got celebrated as proof that AI can replace entire companies. But as one analyst pointed out the billion-dollar narrative is doing more work than the actual business model. Gallagher didn't build something new. He picked a market with desperate demand and a regulatory window that could close at any time, then used AI to fabricate the trust layer on top: fake doctors, fake patients, fake results. The company's own website fine print admits that people in its ads may be actors or AI and not licensed professionals. That's not an AI success story. That's a lesson in how fast AI can scale deception when nobody's looking closely enough.

Evolving AI: Studies from MIT and Stanford reveal AI chatbots agree with users far too often, even when they're clearly wrong.

Two major studies have exposed a troubling pattern in how AI chatbots interact with their users. MIT researchers built simulations of 10,000 conversations and found that even small amounts of AI agreement caused logical people to spiral into extreme confidence about false beliefs. The Stanford study went further by testing 11 popular models including ChatGPT, Claude, Gemini and DeepSeek against nearly 12,000 real-life scenarios where the user was clearly in the wrong. Every model agreed with users roughly 49% more than real humans would. In experiments with over 2,400 real people the results were clear. Those who received agreeable AI responses felt more justified in their actions and less motivated to take responsibility. The researchers describe this as sycophancy — AI flattering users to the point of reinforcing harmful thinking rather than offering honest pushback.

Here's the tricky part: users actually prefer the flattery and rate sycophantic responses as more trustworthy, which gives AI companies zero incentive to fix it. Researchers have already documented close to 300 cases of "AI psychosis" linked to chatbot interactions, and about 12% of US teenagers now turn to chatbots for emotional support. Meanwhile OpenAI had to roll back a ChatGPT update last year after it became so agreeable it endorsed dangerous ideas, and one anthropology professor has gone as far as calling sycophancy a "dark pattern" designed to keep people hooked. The real question is whether these companies can afford to make their products less likeable in order to make them more honest.

Unlock a focused set of AI strategies built to streamline your work and maximize impact. This guide delivers the practical tactics and tools marketers need to start seeing results right away:

Stay ahead of the curve with these top strategies AI helped develop for marketers, built for real-world results.

Download the Free Report

Evolving AI: OpenAI acquired TBPN, a popular tech talk show, marking its first media company purchase.

TBPN airs daily on YouTube and X for three hours covering tech, business, AI and defense. It has built a cult following in Silicon Valley as a space where CEOs like Zuckerberg and Nadella speak candidly with fellow insiders. OpenAI plans to scale the brand while tapping the founders' marketing instincts across the company. The show will sit under OpenAI's strategy team. Altman called TBPN his favorite tech show and said he doesn't expect the hosts to go easier on OpenAI after the deal.

OpenAI isn't just buying a podcast. It's buying influence. And it's part of a much bigger pattern right now. The Ellison family is in the process of buying CNN through their $111 billion Paramount-Warner deal. Jamie Dimon has been floating his own media venture. And now OpenAI, an $852 billion company gearing up for one of the biggest IPOs in history, just bought the talk show its own CEO calls his favorite. "Editorial independence" sounds nice on paper, but the show will report directly to a political strategist who has spent the past two years shaping AI policy in Washington. When the companies building the most powerful technology on Earth start owning the outlets that cover them, the question isn't whether the coverage changes overnight. It's whether anyone will notice when it slowly does.

You think 4x faster than you type. So why are you typing prompts?

Wispr Flow turns your voice into ready-to-paste text inside any AI tool. Speak naturally - include "um"s, tangents, half-finished thoughts - and Flow cleans everything up. You get polished, detailed prompts without touching a keyboard.

Developers use Flow to give coding agents the context they actually need. Researchers use it to describe experiments in full detail. Everyone uses it to stop bottlenecking their AI workflows.

89% of messages sent with zero edits. Millions of users worldwide. Available on Mac, Windows, iPhone, and now Android (free and unlimited on Android during launch).

Try Wispr Flow free →

QUICK HITS

🤠 It’s a Wild West: AI watchdogs say facial recognition policing errors are on the rise.

🧬 Meet MaxToki: The AI that predicts how your cells age and what to do about it.

🎬 Netflix AI team just open-sourced VOID: an AI model that erases objects from videos, physics and all.

🎥 AI is rewiring the world’s most prolific film industry.

 *partner link