DiscoverCyber Security DistrictThe AI Deepfake Problem Is Worse Than You Think | Marco Ramilli | Cyber Security District
The AI Deepfake Problem Is Worse Than You Think | Marco Ramilli | Cyber Security District

The AI Deepfake Problem Is Worse Than You Think | Marco Ramilli | Cyber Security District

Update: 2025-12-10
Share

Description

In this episode of Cyber Security District podcast, we speak with Marco Ramilli, a cybersecurity visionaire, TEDx speaker, and founder of IdentifAI, a deepfake detection company working at the intersection of AI, trust, and digital identity. From writing early sandboxing papers to leading cyber defense teams and selling his previous startup to VirusTotal (Google), Marco has spent the last two decades safeguarding digital borders.


At IdentifAI, he’s now taking on one of the biggest threats of the AI era: manipulated content. Whether it’s fake insurance claims, fraud in football scouting, or synthetic identities onboarding to banks, his platform uses AI to detect AI: verifying whether images, voices, or videos were generated by humans or machines.


In this episode, Marco shares his founder story, why he believes the shift from deterministic to probabilistic computing will reshape cybersecurity, and how his latest startup scaled to enterprise use cases within a year.


In this episode, we cover:

  • Marco’s hacking origin story and first run-in with his university

  • Researching malware and voting systems at UC Davis

  • Founding his first cybersecurity company in Italy when the market wasn’t ready

  • Bootstrapping a business through 3+ years of break-even before momentum hit

  • Early ransomware response and reverse-engineering threats

  • Getting acquired by Tinexta and integrating into VirusTotal

  • Building IdentifAI to detect deepfakes using AI-powered forensics

  • From onboarding fraud to football scouting scams,  real use cases

  • The rise of “digital performers” and the ethics of synthetic content

  • Why deterministic cybersecurity tools are no longer enough

  • The philosophical shift security teams must make to handle probabilistic models

  • Advice for scaling startups, hiring A-players, and letting go as a founder

  • His signal message to CISOs: curiosity will save us


Timestamps:

00:00 – Intro

00:14 – Meet Marco Ramilli

01:00 – Hacking the campus network (with a for loop!)

02:15 – Getting caught, and hired for cyber research

03:10 – Reverse engineering malware at UC Davis

04:30 – Cybersecurity in Italy: too early, too expensive

05:45 – Why compliance was seen as a cost

06:20 – Building a company just to be the “owner of his time”

07:00 – Early business struggles & staying focused

08:30 – Why saying “no” saved his startup

10:10 – On embracing failure and learning from loss

11:30 – Curiosity, discomfort, and moving to the U.S.

13:00 – Sleeping on floors and working without backup

14:00 – Founding Yoroi, growing from 4 to hundreds of employees

15:30 – The ransomware era and massive inbound traction

16:45 – Sandboxing: research, papers, and real-world applications

18:00 – Joining Google via acquisition

19:30 – Startup phases and letting go as a founder

21:00 – Hiring advice: don’t save money on great people

23:00 – Launching IdentifAI: how a fake Pope jacket sparked a real mission

25:00 – Building AI to detect AI: early model design

27:00 – From 80% to 96% accuracy in deepfake detection

28:00 – Why images are more dangerous than text

29:45 – The weaponization of synthetic media

31:20 – How IdentifAI detects voice, image, and video manipulation

32:45 – Use cases: banking KYC, insurance fraud, and football scouting

35:00 – API-first strategy and enterprise readiness

36:30 – “Digital Performers” and the ethics of representation

37:30 – Real-time meeting detection via agents

39:00 – Why probabilistic AI is changing the rules

40:00 – How cybersecurity must evolve to deal with uncertainty

43:20 – Deployment flexibility: on-prem or private cloud

46:20 – The philosophical challenge of non-deterministic systems

49:30 – Final message to CISOs: stay curious


Connect with the guests:



Follow Cybersecurity District:


Comments 
In Channel
loading
00:00
00:00
x

0.5x

0.8x

1.0x

1.25x

1.5x

2.0x

3.0x

Sleep Timer

Off

End of Episode

5 Minutes

10 Minutes

15 Minutes

30 Minutes

45 Minutes

60 Minutes

120 Minutes

The AI Deepfake Problem Is Worse Than You Think | Marco Ramilli | Cyber Security District

The AI Deepfake Problem Is Worse Than You Think | Marco Ramilli | Cyber Security District

Cyber Security District