Friends, today yet another media outlet is here to perform "AI doomsday literature" for us.

Here's what actually happened: a team from Stanford and the ARC Institute published a paper in Nature Biotechnology saying they used a genome language model to design bacteriophage viruses from scratch — out of 302 candidates, 16 successfully achieved self-replication. The research goal is clear: to fight antibiotic-resistant bacteria, to develop phage therapy. Put simply: use AI to design viruses that only kill bacteria, to save people infected with superbugs that antibiotics can no longer stop.

This is obviously a good thing, right? Guess how some media outlets headlined it?

"Shocking! AI Designs Virus for the First Time — Humanity Opens Pandora's Box!"

"AI Creates Supervirus, Biosecurity Red Light Flashes Again!"

"Creating Life from Scratch: AI Writes Viral Genomes — How Far Is Bio-Crisis?"

I almost thought AI had created a T-virus that could wipe out humanity. Then I looked at the paper — oh, bacteriophages. Bacteriophages — the kind of virus that only infects bacteria, not human cells. They exist to kill bacteria, and they've been around for billions of years in nature, more numerous than bacteria themselves. Take a sip of water and there are hundreds of millions of phages in it.

The Classic Three-Axe Hype Playbook, All Present and Accounted For

Friends, I've seen this clickbait formula in AI biology news way too many times — it's the classic three-axe routine.

Axe one: swap the concept, replace 'bacteriophage' with 'virus.' Regular people hear "virus" and immediately think COVID, HIV, Ebola — instant flashback to "deadly pandemic" fears. But what's a bacteriophage? It's a virus of bacteria. It couldn't care less about human cells. It's like saying "scientists design a new pesticide" and the media headline reads "scientists create deadly poison" — not technically wrong, but the entire vibe is skewed.

Axe two: bury the context, never mention it's for treating disease. Why research phage therapy? Because antibiotic resistance has become a global public health crisis — millions of people die every year from drug-resistant bacterial infections, and existing antibiotics are working less and less. Phage therapy is one of the most promising alternative approaches. AI accelerating phage design essentially means accelerating the development of life-saving medicine. But the media doesn't care — they just tell you "AI made a virus" and don't tell you it's made to save lives.

Axe three: "creating life from scratch," leaning into the grand narrative of "the origin of life." Stuff like "AI designs life for the first time," "humanity playing God" — made to sound all mystical and profound. Leaving aside whether viruses even count as life (a debate still raging in the scientific community), even if they do — this isn't designing a new species from zero. It's using language models to design and optimize based on known phages, then synthesizing, validating, and screening them in a lab. Only 16 out of 302 candidates worked — less than 5% success rate. That level is still leagues away from "playing God."

The Real Risks Get Buried

To be clear, I'm not denying the potential risks of AI synthetic biology — risks exist, and they need to be taken seriously. But the risk isn't in the AI model itself; it's in lab management and regulatory frameworks.

Think about it. A malicious person can synthesize dangerous things with or without AI — gene synthesis technology is already quite mature. If you have money, you can find a contract company to synthesize whatever gene sequence you want. It's true that AI lowers the barrier to design, but the biggest bottleneck was never design — it's synthesis and experimental conditions. And phage therapy research is public and regulated; its purpose is to cure and save people — that's a completely different category from "AI building bioweapons."

What actually deserves vigilance? It's that the barrier to entry for biological experiments is getting lower and lower, that regulation of gene synthesis services isn't robust enough, that international rules for dual-use research remain unclear. These are the real biosecurity issues — not problems with any AI model. Blaming it on AI instead makes everyone ignore the problems that actually need solving.

But the media could care less. "AI designs phage therapy to fight antibiotic resistance" — who clicks on that title? "AI creates supervirus, bio-crisis imminent" — click-through rate goes straight through the roof. The readers get scared, the KPIs go up, and whether the public's understanding of AI and biotechnology gets distorted? Who cares?

Friends, next time you see headlines like "AI creates life," "AI builds a virus," "AI designs doomsday weapons" — take three deep breaths, then ask yourself three questions: first, what does this thing infect? Bacteria or humans? Second, what was it originally made for — healing or harming? Third, what's the success rate — 100% or 5%?

After asking those three questions, you'll realize: the vast majority of "AI doomsday news" is just KPI wrapped in fear.

That's all the roasting for today. More tomorrow.