Why Autonomous AI Agents Fail in Production (And How to Fix It) | Bangalore Tech Meetup

Learn about the hidden security vulnerabilities in popular agent protocols (A2A, MCP, ACNP), how to prevent agent goal hijacking, and the "Passport & Visa" framework for governing autonomous agent interactions in production. 🔔 Subscribe to stay ahead in enterprise AI, open architectures, and data strategy:    / @clouderainc   🔹 Key Topics Covered: The 3 Core Questions: Contain, Forecast, and Adjudicate Agent Security Protocol Analysis: Transport, Prompt Injection, and Side-Channel Privacy Leakage The "Passport & Visa" Framework for Agentic AI Containment Token Budget Forecasting & Probabilistic Failure Scoring Live Demo: Ungoverned vs. Governed vs. Disputed AI Agents Join the Cloudera Community to learn more! 👉 https://community.cloudera.com Full Playlist:    • Future of Data Meetup Livestreams   Get Started 🔹 Request a Demo: https://www.cloudera.com/products/clo... 🔹 Customer Success Stories: https://www.cloudera.com/customers.ht... 🔹 Read the Cloudera blog: https://www.cloudera.com/blog.html Resources: Hear more from our customers:    • Customer Stories   Watch Demos:    • Demos   Chapters: 00:00 Welcome and Intro 05:10 Contain Forecast Adjudicate 27:57 DEMO Governed vs Disputed Agents 33:10 QA Token Budgets and Costs 48:50 Product Catalog Architecture 53:00 Analyzing Vulnerabilities with Docker Scout 1:16:00 Docker Hardened Images Deep Dive 1:35:00 Local Sandboxing with MicroVMs 1:51:00 DEMO MicroVM and MCP Sandboxing 2:18:00 What is a Sandbox Kit 2:28:00 DEMO AWS Environment Simulation 2:48:00 Why We Dont Trust AI Models 3:00:00 Building Trust Models and Evidence 3:13:00 Closing and Careers Connect with Cloudera LinkedIn ►   / cloudera   Facebook ►   / cloudera   X ►   / cloudera   Spotify ► https://open.spotify.com/show/102S8zo... #AIGovernance #AgenticAI #AIAgents #CyberSecurity #OpenSource #Cloudera #TechMeetup