Recent breakthroughs in artificial intelligence (AI) are revolutionizing several areas and integrating into daily life. Large language models (LLMs) are a significant part of this transformation and reducing human intervention. Recent research shows that agentic AI is the next breakthrough, which can operate independently and make decisions without human involvement. This chapter provides a comprehensive understanding of the security threats, risks, and challenges related to agentic AI. Thus, when each of the components of the AI agents work flow is reduced to its first principles, we are in a position to evaluate the risks and vulnerabilities as the research progress. In this way, this study fills the current shortage of literature and combines knowledge available and serves as a background for further research in this field. This chapter further represents the findings with the ongoing work of the OWASP agentic AI, which is building one such top 10 for these systems. The primary contribution of this chapter lies in understanding the distinctive security challenges of agentic AI, proposing possible solutions, and presenting the adaptive secure agent framework (ASAF) as a comprehensive blueprint for establishing robust security norms. This research is critically important for stakeholders who wish to leverage the powerful capabilities of agentic AI while ensuring effective protection against emerging risks.