AI systems can feel like black boxes. But when you test them like any other software, surprising weaknesses start to surface.
What happens when an LLM is pushed beyond its intended boundaries?
Can it expose sensitive data, bypass safeguards, or behave in ways you did not anticipate?
In this session, we explore how the principles behind OWASP and its widely recognized security guidance can help you better understand and mitigate risks in LLM-powered applications.
This is not a theoretical lecture. You will see real-world examples drawn from news stories and ongoing projects, showing how vulnerabilities in AI systems actually play out. Through practical demonstrations and original cartoons, complex risks are broken down into clear, memorable lessons.
What You’ll Learn
Large Language Models introduce new attack surfaces, but many of the risks are more familiar than they seem. In this session, we cover:
- How poorly crafted prompts can become entry points for attackers
- How weak safeguards allow sensitive information to leak
- How biased training data can create hidden security exposure
- How lack of logging and monitoring makes AI failures harder to detect
- How traditional testing skills remain highly effective for AI security
You will walk away with practical strategies to strengthen your LLM integrations and a structured way to think about AI risks.
Why This Matters
Security issues in AI systems rarely appear dramatic at first. Small gaps, unclear boundaries, or misplaced trust in model behavior can quietly grow into serious problems.
Applying structured security thinking and established risk models allows teams to move from reactive fixes to proactive prevention.
Key Takeaways
- AI security is continuous: LLM risks evolve as attackers adapt. Ongoing testing and review are essential.
- Traditional testing still works: Exploratory testing, risk analysis, and structured thinking remain highly effective.
- System awareness is critical: Understanding how AI interacts with the rest of your application helps close security gaps.
Watch the Recording
Missed the live session or want to revisit the insights?
Watch the full recording to gain practical tools, a clearer risk framework, and a grounded approach to securing LLM applications.
You do not need new tools to improve AI security. You need the right mindset and a structured approach.
