Mikhail Zolotarev

From threat modeling autonomous AI agents to exploring new forms of prompt injection, Mikhail brings an offensive security perspective to the world of generative AI. With a foundation in application and infrastructure penetration testing — especially in high-assurance environments like the financial sector — he now focuses on exposing real-world weaknesses in the way AI systems are built and deployed. His recent work includes researching the security posture of Model Context Protocol (MCP) infrastructures, evaluating the practical impact of prompt injection attacks, and measuring the effectiveness of current mitigation strategies. Mikhail also contributes to the development of internal methodologies for AI security testing, combining traditional offensive techniques with emerging AI threats, helping teams assess LLM-integrated applications beyond the surface level and offering a grounded and actionable view into where these technologies are vulnerable — and how to test them effectively.