This talk covers

• 1) Evolution of GenAI and Risks Involved
• 2) Introduction to prompt Hacking(with Examples): Food Service exploit (Uber Eats), Google Docs Exploit (Rephrase Document feature), Skip the Queue exploit - ChatGPT 4 registration
• 3) Prompt Injection Attack Scenario
• 3.1) Direct
• 3.1.1) Direct Prompt Injection: ChatGPT 4
• 3.1.2)Direct Prompt Injection: Information Disclosure Vulnerability in Healthcare Application with Threat Model and Threat Statement
• 3.2) Indirect prompt Injection
• 3.2.1) Microsoft Bing - Data Exfiltration
• 4) Mitigations - Prompt Segmentation and Privilege Boundaries
• 5) Key Takeaways and Wrap Up