Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
This Black Hat conference talk demonstrates advanced prompt injection exploits targeting major LLM applications including Microsoft Copilot, Google Gemini, NotebookLM, Apple Intelligence, GitHub Copilot Chat, and Anthropic Claude. Discover real-world examples of sophisticated threats including misinformation techniques using conditional instructions, automatic tool invocation exploits that escalate privileges and extract sensitive data, data exfiltration methods using markdown and hidden payloads to bypass security controls, SpAIware techniques for maintaining long-term control through LLM memory manipulation, and ASCII smuggling methods that hide text from users. Learn about mitigations for each vulnerability category and how vendors are addressing these security challenges. Presented by Johann Rehberger, Security Researcher and Red Team Director at embracethered.com, this 39-minute presentation provides essential insights for understanding the security implications of large language model applications.
