Breaking & Securing AI Agents: MCP Risks, Prompt Injection & Azure Defenses | Mani Bindra

AI agents are powerful - but how do you secure them against real-world threats?

In this session from **GitHub Copilot Dev Days - Mumbai (#Apr26MTP)**, Mani Bindra explores the security challenges of modern AI systems and demonstrates how to design safer, production-ready architectures.

This is a demo-heavy session focused on real risks and practical defenses:

✅ Prompt injection attacks and how they work
✅ MCP (Model Context Protocol) risks and attack surface
✅ Tool poisoning and agent misuse scenarios
✅ Isolation strategies and sandboxing approaches
✅ Secure Azure architectures for AI agents
✅ Building guardrails for production-grade AI systems

---

📍 **Event:** GitHub Copilot Dev Days – Mumbai
🏢 **Venue:** Microsoft Mumbai
📅 **Date:** 18 April 2026

---

🙏 Special thanks to Mani Bindra for sharing a deeply practical and security-focused session with the community.

---

🔗 **Explore more from the event:**
📸 Photo Gallery: https://hackmum.in/past-events/april-2026-github-copilot-dev-days/gallery/
📚 Past Event Resources: https://hackmum.in/past-events/

---

💙 Organized by:
Mumbai Technology Meetup / Hackerspace Mumbai (#mumtechup)

---

📢 Stay connected for upcoming meetups:
🐦 https://x.com/hackmum

---

#GitHubCopilotDevDays #AISecurity #MCP #PromptInjection #Azure #Apr26MTP #mumtechup

Видео Breaking & Securing AI Agents: MCP Risks, Prompt Injection & Azure Defenses | Mani Bindra канала Hackerspace Mumbai