​Most teams building AI applications spend their time making the AI more capable.

​Better prompts. Better models. Better workflows.

​But capability isn't the only thing that matters.

​What happens when someone intentionally tries to break your AI system?

​A user convinces your chatbot to ignore its instructions.

​An AI assistant reveals information it was never supposed to share.

​A carefully crafted prompt causes an AI agent to behave in ways its creators never intended.

​These aren't hypothetical scenarios.

​They're happening every day across AI products.

​As organizations increasingly deploy AI into customer-facing and business-critical workflows, understanding how these attacks work is becoming just as important as building the AI itself.

​In this session, we'll explore how prompt hacking works, why AI systems are vulnerable to it, and how builders can identify and defend against these attacks.

​Using real-world examples and live demonstrations, we'll walk through the mindset, techniques, and tools used to test AI systems safely. We'll use Gandalf by Lakera, a prompt-hacking playground used by over a million people, to demonstrate common attack patterns in a controlled environment.

​We'll also explore AI bug bounty and red-teaming platforms where researchers are invited to test AI systems with permission, and discuss how organizations are building security programs around AI applications.

​The goal isn't to teach people how to attack random AI systems.

​It's to help builders understand how these vulnerabilities work so they can build more secure AI products.

​What We'll Cover

• ​Why prompt hacking matters

• ​Prompt engineering vs. prompt hacking

• ​Common prompt hacking techniques and attack patterns

• ​A mental model for discovering prompt vulnerabilities

• ​Live demonstrations using Gandalf by Lakera

• ​AI red teaming and bug bounty platforms such as Prompt2Hack

• ​Responsible AI security testing and disclosure

• ​Real-world AI bounty programs including Microsoft Copilot and HackerOne initiatives

• ​Defending AI applications against prompt attacks

• ​Useful security tools, frameworks, and libraries for AI builders

​Why You Should Attend

• ​You're building AI products or internal AI tools

• ​You want to understand how AI systems get attacked

• ​You're curious about prompt injection, jailbreaks, and AI security

• ​You want practical ways to test and secure AI applications

• ​You're exploring AI agents, copilots, or LLM-powered workflows

• ​You want real-world examples instead of theoretical security discussions

​What You'll Walk Away With

• ​A clear understanding of prompt hacking fundamentals

• ​A framework for identifying vulnerabilities in AI systems

• ​Practical techniques for testing AI applications safely

• ​Best practices for defending AI products against common attacks

• ​A list of tools and resources for AI security testing

• ​A better understanding of responsible AI red teaming

​Who This Is For

• ​AI Engineers

• ​Developers building LLM applications

• ​Product Managers

• ​Security Engineers

• ​AI Researchers

• ​Founders building AI-native products

• ​Anyone interested in AI security

​Join us to learn how attackers think, how AI systems get exploited, and how to build AI applications that are far more resilient in the real world.

​Join our WhatsApp community for the latest AI news, tools, hacks, and mental models: https://chat.whatsapp.com/CSg6OChjPmCHt2lASGJKgA

​Explore other resources at https://www.aiatflytbase.com/