Breaking Through in Cybersecurity Marketing 6.3.26
Ep 214 | 6.3.26

What AI Security Research Actually Looks Like with John Zenick of Harmonic Security

Show Notes

Episode Summary:

John Zenick started his cybersecurity journey modding a Nintendo Wii in middle school. He is now an AI Security Researcher at Harmonic Security and a Teaching Fellow at Harvard, and joins our show to discuss everything AI! Even though we're a marketing podcast, of course we love to talk to the people doing the work in AI security today.

John explains what shadow AI looks like inside organizations, why blocking it rarely works, and where the real data risk sits. He also brings out a pocket-sized device he built that runs a local LLM with no internet connection, and makes a case for why your process may be just as worth protecting as your data.

Chapters:

[00:00] Introduction to Cybersecurity Marketing and AI Research

[01:10] John Zenick's Journey into Cybersecurity

[02:57] From Hacking to Home Networking

[04:15] Academic Pursuits and Teaching Cybersecurity

[07:09] Current Role at Harmonic Security

[09:52] AI Security Research and Its Challenges

[11:49] Understanding Shadow AI and Its Implications

[13:41] The Future of AI Tools and Data Privacy

[15:40] Exploring Open Source AI Models

[18:12] Building Private AI Models for Enhanced Privacy

[24:50] Innovative AI Devices and Their Applications

[31:19] Resources for Building Your Own AI Models

[32:46] Personal Insights and Future Aspirations

About John:

John Zenick is an AI Security Researcher at Harmonic Security and a Teaching Fellow at Harvard University, where he has taught Software Engineering with Generative AI and supports applied computation capstone projects. His background spans incident response, security research, and detection engineering. He has also built local AI infrastructure from scratch, including private LLM systems designed to run fully offline.

Links & Resources:

John Zenick on LinkedIn

Harmonic Security

Ollama - for getting started with local open source models

Hugging Face - open source models and datasets

John'sHugging Face profile - quantization scripts and models for edge devices

Andrej Karpathy on YouTube - recommended for learning how LLMs work

Subscribe & Review:

👉 Enjoyed this one? Drop us a quick ⭐ rating or review, it helps more cyber marketers find the show.

📩 Questions, feedback, or want to be a guest? Email: podcast@cybersecuritymarketingsociety.com 

Follow the Host:

Thank you for tuning in to another episode of Breaking Through in Cybersecurity Marketing. You can join the Cybersecurity Marketing Society on our website, on our main LinkedIn page, our Instagram page, or our podcast LinkedIn page. 

See you on the next episode!