Fighting deepfakes, and using bacteria to deliver medicine inside the body

Science Magazine Podcast31mMay 14, 2026

Get the full intelligence

Search transcripts, export clips, track mentions, and explore all topics from “Fighting deepfakes, and using bacteria to deliver medicine inside the body” inside PodZeus.

AI-Generated Summary

This episode of the Science Magazine Podcast explores two cutting-edge scientific frontiers: the escalating threat of AI-generated deepfakes and the innovative use of engineered bacteria for targeted medical treatments. Host Sarah Crespi begins with a segment on digital forensics expert Hany Farid, who is at the forefront of detecting manipulated media. Farid explains how AI-generated videos often fail in physical realism—such as incorrect sound-to-video timing, implausible physics, and inconsistent shadows—making them detectable through rigorous analysis. Despite the growing volume of fake content, he remains concerned about the societal erosion of trust in visual evidence and the difficulty of scaling human verification. The episode then shifts to a conversation with Tetsuhiro Harimoto, a researcher developing encapsulated, genetically engineered bacteria that can sense infections and deliver targeted therapies inside the body. His team created a robust, non-brittle hydrogel capsule capable of withstanding the immense mechanical pressure generated by rapidly dividing bacteria. In mouse models, these living implants successfully reduced antibiotic-resistant infections around medical implants and showed promise for future applications in cancer therapy and smart, responsive medical devices. The episode underscores both the urgent need for truth verification in the digital age and the transformative potential of living medicine.

Key Takeaways
1

AI-generated deepfakes are becoming more convincing, but physical inconsistencies like sound-video misalignment and incorrect physics can still reveal manipulation.

2

Human-led media verification is overwhelmed; scalable automated systems are needed, but they risk becoming part of the same problem.

3

Engineered bacteria encapsulated in specially designed hydrogels can survive inside the body and respond to infections autonomously.

4

The mechanical force generated by rapidly dividing bacteria requires materials strong enough to contain them without being brittle.

5

Living implants could revolutionize medicine by enabling continuous, adaptive, and localized drug delivery for infections and cancer.

…and 1 more takeaway available in PodZeus

Chapters
0:00
9 min

The Deepfake Arms Race and the Erosion of Visual Truth

We're living through a moment in time where something that we've always just kind of taken for granted is eroding, and that's been... the ability to say whether an image is real or not.

Highlight
9:00
16 min

How Physics and Perception Reveal AI Manipulation

Farid explains how he uses principles of physics and photogrammetry to detect fake videos. He demonstrates how AI struggles to replicate real-world physical laws—such as the parabolic trajectory of a thrown ball or the correct delay between sound and visual effects. Humans are poor at detecting these inconsistencies because our perception evolved to ignore them. The episode underscores that while AI may one day master these details, the current gap offers a window for detection.

25:00
15 min

Engineering Living Medicine: Bacteria as Smart Drug Delivery Systems

What if we could make these implantable devices as a living entity that can sense and produce drugs continuously and even evolve over time or something? That would be really exciting.

Highlight
40:00
11 min

From Lab to Body: Testing and Future of Living Implants

Harimoto details the experimental validation of the bacterial capsules, including long-term culture tests and implantation in mice. The capsules remained intact for six months and effectively suppressed antibiotic-resistant infections. The technology is designed to be explanted with the medical implant, avoiding long-term risks. The team envisions broader applications, including cancer therapy and self-regulating medical devices that adapt to disease states.

High-Impact Quotes
What if we could make these implantable devices as a living entity that can sense and produce drugs continuously and even evolve over time or something? That would be really exciting.
Tetsuhiro Harimoto29:18
Viral: 90.0
We're living through a moment in time where something that we've always just kind of taken for granted is eroding, and that's been the ability to say whether an image is real or not.
Hany Farid1:44
Viral: 85.0
If you try to AI generate that, it'll do something arced. But will it move in precisely a parabola? There's an initial velocity and there's gravity. So that means if you look at that same time lapse, right, the gap will get smaller and smaller and then increase, increase when gravity takes over.
Hany Farid7:24
Viral: 75.0
Speakers

Host

Sarah Crespi

Guests

Hany FaridTetsuhiro Harimoto
Topics Discussed
AI-generated deepfakes95%Living medicine and engineered bacteria92%Digital forensics and media verification90%Encapsulated microbial therapeutics88%Physics-based detection of fake media85%Biocompatible materials and hydrogels80%Synthetic biology in medicine78%Smart medical implants75%
People & Brands

Sarah Crespi

person

18xNeutral

Hany Farid

person

14xPositive

Tetsuhiro Harimoto

person

12xPositive

E. coli

other

4xNeutral

Icahn School of Medicine at Mount Sinai

organization

3xPositive

Science Magazine

other

3xPositive

Paramount Plus

other

2xNeutral

Polyvinyl alcohol (PVA)

other

2xNeutral

AAAS

organization

2xNeutral

Kai Kupfer-Schmidt

person

2xNeutral

Get the full intelligence

Search transcripts, export clips, track mentions, and explore all topics from “Fighting deepfakes, and using bacteria to deliver medicine inside the body” inside PodZeus.

Start discovering podcast insights today

Start with a 7-day trial and explore a growing catalog of popular podcasts. No credit card required.

No credit card required • 7-day trial • Cancel anytime