Скачать или смотреть Eito Miyamura Reveals How He CRACKED ChatGPT's AI Security w/ Prompt Injection

📲 Work with me 1:1: https://cal.com/georgesiosi/coaching-ai
🗞️ Subscribe to newsletter: https://georgesiosiai.substack.com
👨‍🔬 Join my AI.GSD Lab: https://georgesiosi.com/gsdlab

--------

In this episode, join ‪@georgesiosi‬ and Eito Miamura as they dive into the critical issues surrounding AI security, focusing on a recent experiment where Eito and his team were able to manipulate ChatGPT to leak private email data.

Eito, founder of Edison.Watch, explains the various attack vectors and the urgent need for robust AI firewalls. The discussion covers the emerging threats of AI-enabled social engineering, the responsibilities of developers, and how organizations can protect themselves from these risks.

Get insights into the necessary precautions for using MCPs (Model Context Protocol) and the future of AI safety in an increasingly complex world.

⏰ Chapters:
00:00 Intro & Guest Welcome
00:14 The Tweet That Sparked the Conversation
00:54 Understanding AI Security & Data Exfiltration
01:50 Simple Yet Dangerous AI Attacks
03:28 Emerging AI Jailbreaks and Their Implications
04:46 Real-World Examples of AI Exploits
06:16 Current State and Future of AI Security
07:06 Social Engineering and Developer Risks
13:11 Edison Watch: Mission and Team
22:17 Final Thoughts and Call to Action

🔗 Links:
Edison Watch (Eito's startup): https://edison.watch
Open Edison: https://github.com/Edison-Watch/open-...
Eito's X profile: https://x.com/Eito_Miyamura (connect with Eito)