Join us for a critical examination of OpenAI's GPT-4.1 models as we unveil vulnerabilities discovered through sophisticated red teaming techniques. Diana Kelley, CISO at Protect AI, sits down with Sailesh Mishra, an AI security expert and Recon Product Leader at Protect AI, to unpack the recent security assessment findings on the GPT-4.1 series (including Nano and Mini variants).This eye-opening discussion explores how these advanced models with extended context windows perform against targeted security testing—revealing medium-risk ratings across the board with notable vulnerabilities.
What You'll Learn:
- Understanding GPT-4.1 Models - Discover the key differences between reasoning models and long context window models, and why these distinctions matter for enterprise security
- AI Red Teaming Methodology - Learn how Protect AI's Recon tool systematically evaluates LLM safety alignment using agent-based simulations and first-party attack libraries
- Critical Vulnerabilities Exposed - Explore how prompt injections and evasion techniques remain significant threats, with GPT-4.1 Nano showing particular susceptibility
- Real-World Attack Scenarios - See demonstrations of how these models can be manipulated to generate harmful content including malicious code and cyberattack instructions
- Enterprise Risk Mitigation - Gain actionable insights on continuous monitoring, model selection strategies, and architectural approaches to reduce AI security risks
Who Should Watch:
Security leaders, AI practitioners, application developers, and anyone integrating or deploying OpenAI's models in production environments.
Diana Kelley, Chief Information Security Officer
Sailesh Mishra, Recon Product Leader