The purpose of Vulnerable AI Factory is to provide a safe, hands-on playground for learning and experimenting with security vulnerabilities in modern agentic AI systems. By simulating real-world risks in components such as MCP (Model Context Protocol) server and clients, agent-to-agent protocol (A2A) integrations, vector databases, and autonomous pipelines, the project allows developers, security researchers, and students to explore, test, and understand attacks like prompt injections, model misconfigurations, and adversarial manipulations - turning potential AI failures into practical learning opportunities.

Road Map

During its first year, Vulnerable AI Factory will focus on establishing a robust foundation for hands-on agentic AI security training and community engagement:

Project Setup and Initial Release:

#* Develop and release the initial version of the Vulnerable AI Factory playground. #* Provide a set of curated challenge scenarios and sandbox experiments for users to safely explore AI security vulnerabilities.

Documentation and Learning Resources:

#* Create comprehensive guides, tutorials, and step-by-step exercises to help users understand AI security concepts, attack vectors, and defensive strategies.

Community Building and Outreach:

#* Establish a presence within the OWASP community and broader AI/security ecosystem. #* Encourage contributions from developers, researchers, and educators to expand challenges, experiments, and AI components.

Feedback and Iteration:

#* Gather user feedback and contributions to refine playground features, improve challenge quality, and extend the scope of agentic AI components.

Demonstrations and Awareness:

#* Present the project at (OWASP) conferences, workshops, or webinars to raise awareness of AI security risks and promote hands-on learning.

By the end of its first year, Vulnerable AI Factory aims to be a stable, extensible, and community-supported platform for exploring and learning about agentic AI security, laying the groundwork for long-term growth and broader adoption.