Hijacking AI Memory: Inside Johann Rehberger's ChatGPT Security Breakthrough

Secure Talk Podcast

2025/04/01

Overview Shownote Highlights Transcript Chapters Pins

Shownote

In this eye-opening episode of SecureTalk, host Justin Beals interviews Johann Rehberger, a seasoned cybersecurity expert and Red Team Director at Electronic Arts, about his groundbreaking discovery of a critical vulnerability in ChatGPT's memory system. ...

Highlights

In this episode of SecureTalk, host Justin Beals sits down with Johann Rehberger, a leading cybersecurity expert and Red Team Director at Electronic Arts, to explore the emerging security risks in artificial intelligence systems. Drawing from his extensive background in cybersecurity, Johann delves into how traditional security principles are being challenged by the rise of AI technologies like ChatGPT.

00:03

Johann discovered a critical vulnerability in ChatGPT's memory using SPAIWARE.

23:55

SPAIWARE allowed persistent prompt injection to leak chat history via invisible image tags.

35:50

A bypass was found to hijack ChatGPT and steal data through prompt injection.

38:37

Prompt injection is now recognized in bug bounty programs.

Chapters

From Web Security to AI: A Cybersecurity Expert’s Journey

00:00

Uncovering ChatGPT’s Memory Vulnerability and the Birth of SPAIWARE

15:11

Prompt Injection: Weaponizing AI Through Remote Control Techniques

26:46

Securing the Future: Ethical Hacking and the Challenge of Autonomous AI

38:37

Transcript

Justin Beals: Hello, and welcome to SecureTalk, a podcast where we explore the critical world of information security, innovation and compliance. I'm your host, Justin Beals, founder and CEO of StrikeGraph. Together with our expert guests, we'll provide yo...