Securing AI Agents in Production: Mexico City

A two-day, hands-on workshop for the teams shipping AI agents to customers. Map the 10 attack surfaces, test against the OWASP Top 10 for LLMs, and build runtime controls and kill switches. Delivered in English.

Overview

AI agents now hold credentials, call APIs and act on customer data. For SaaS companies and growing businesses, that is also a sales question: every enterprise buyer, auditor and regulator will ask how those agents are secured. Most security programs were not built to answer it.

This two-day workshop gives the people who own security a working method to find, test and contain AI agents. You'll map the 10 attack surfaces of AI systems, work through the OWASP Top 10 for LLM Applications, and walk through recent AI breaches and what stopped them, or didn't. Then the focus moves to control: runtime policy enforcement, incident response for agents, and how to scope kill switches so you can stop one agent, tool or workload without taking everything else down. You'll finish by running a live AI discovery assessment for your own company.

You'll leave with

A map of the 10 attack surfaces across your own AI agents
Runtime controls and scoped kill switches you have built in a live sandbox
An AI discovery assessment of your own company
Clear answers for customer security reviews and audits
A certificate of AI security awareness

Who leads it

Roshan Shaik, founder and CEO of RuntimeAI, with 25+ years in security and product engineering at Okta, Equinix and Cisco. He writes the weekly AI Security Digest. Hosted by Agents Stack, with a guest fireside chat. Vendor-neutral.

Examples focus on financial services, healthcare and agri-produce. Join for both days or day one only. Delivered in English. Bring a laptop: the labs run in a live sandbox.

Agenda

About 12.5 hours of instruction over 2 days.

Day 1: Map the risk

9:00 to 9:30 Welcome and Threat Landscape
9:30 to 10:45 OWASP Top 10 for LLM Applications I
11:00 to 12:30 OWASP Top 10 for LLM Applications II
13:30 to 14:45 OWASP Top 10 for LLM Applications III
15:00 to 16:15 Agentic-Specific Risks
16:15 to 16:30 Day 1 Recap and Q&A

Day 2: Contain it

9:00 to 10:15 AI Governance Frameworks: NIST AI RMF
10:30 to 11:30 The Kill Switch Concept (vendor-neutral)
11:30 to 12:30 Incident Response for AI Agents
13:30 to 15:00 RuntimeAI Session: Live Hands-On
15:15 to 16:15 Capstone Exercise
16:15 to 16:30 Wrap-Up and Next Steps

Who should attend

Security leads at SaaS companies and SMEs shipping AI agents to customers
CISOs, CTOs and engineering leaders who need to answer how their agents are secured
Teams in financial services, healthcare and agri-produce, where agents touch payments, patient data and partner systems first