AMA SessionBuilding Secure AI Products:
The New Threat Model
- August 19th, 2026 10:00 AM EDT
Introduction
AI agents are rapidly becoming part of modern software products. They can search across systems, interact with APIs, execute workflows, and make decisions on behalf of users.
While these capabilities unlock entirely new product experiences, they also change how security vulnerabilities behave. A flaw that once had limited impact in a traditional application can become significantly more dangerous when an AI agent has the ability to discover information, chain multiple actions together, and operate with broad permissions across connected systems.
On August 19th at 10:00 AM EDT, Capria Ventures will host a live session with Akash Mahajan, Founder & CEO of Appsecco and Kloudle, exploring how security changes as products evolve from traditional software into AI-powered and agentic applications.
Rather than focusing on theory, the session will center around a live demonstration. Akash will begin with a familiar application vulnerability, show how the introduction of an AI agent fundamentally changes the attack surface, demonstrate how attackers can exploit these new capabilities, and then walk through the practical techniques engineering teams should use to build secure AI systems.
The discussion will also explore the new security challenges introduced by agentic architectures, including authorization, tool access, MCP security, identity, and runtime controls.
Designed for founders, CTOs, engineering leaders, product teams, and AI practitioners, this session provides practical guidance for anyone building AI-powered products and preparing to move them safely from MVP to production.
What You’ll Learn
Why Agentic Applications Change Security – How adding AI agents to existing products changes the threat model and why familiar vulnerabilities can become significantly more impactful.
Live Security Demonstration – Watch a real application evolve into an agentic system, see how an existing vulnerability can be exploited in new ways, and learn how attackers chain actions through AI agents.
New Attack Surfaces in AI Products – Understand the emerging security risks introduced by AI agents, tool calling, MCP, identity, authorization, and autonomous workflows.
Securing MCP-Based Systems – Best practices for building secure Model Context Protocol (MCP) implementations, including authentication, authorization, least privilege, and safe tool design.
Building Defense Into Agentic Systems – Learn the engineering controls that matter most when deploying AI applications, including permission boundaries, runtime protections, validation layers, and secure system design.
Practical Security Checklist – Leave with a practical checklist your team can immediately apply before moving AI-powered products into production.
Format
Unlike traditional webinars, this is a live AMA conversation:
- No long keynote or slides
- Moderated discussion (Will Poole ↔ Akash Mahajan)
- Live security demonstration
- Interactive audience participation
- 20% of the session dedicated to audience Q&A
Event Details
Title: Building Secure AI Products: The New Threat Model
Speaker: Akash Mahajan
Host: Capria Ventures – GAIN Community
Date: August 19th, 2026
Time: 7:00 AM PT | 10:00 AM EDT | 3:00 PM WAT | 7:30 PM IST | 10:00 PM SGT
Meeting Link: Click here to join the session >
About the Speaker
Akash Mahajan – Founder and CEO of Appsecco and Kloudle | Security for AI SaaS, Agents & MCP
Akash Mahajan is the Founder and CEO of Appsecco and Kloudle, with more than two decades of experience helping organizations identify, exploit, and remediate security vulnerabilities across applications, cloud infrastructure, and modern software systems. Today, his work focuses on one of the fastest-evolving areas of cybersecurity: securing AI agents and agentic applications. His expertise spans application security, Model Context Protocol (MCP), identity and authorization, cloud security, runtime controls, and AI red teaming.
Through Appsecco, Akash leads offensive security assessments of real-world AI systems where autonomous agents interact with APIs, tools, data sources, and cloud infrastructure.
At Kloudle, Akash is building security tools for humans and AI agents across cloud infrastructure, web applications, OAuth and authorization, MCP systems, and endpoints, with a focus on making security testing and investigation directly accessible to agents as well as engineering teams.
Known for his practical, attacker-led approach, Akash demonstrates how systems fail before explaining how to fix them. Rather than focusing on theoretical security concepts, he helps engineering teams understand how AI changes the security landscape and what practical controls they can implement to build secure AI products.
Connect with Akash on LinkedIn to follow his work on AI security, application security, and secure agentic systems.
