WeAreDevelopers Logo WeAreDevelopers
Sign in
For You LIVE
World Congress 2026 North America • Sep 24, 2026 • Session details

I Don't Trust AI Agents (And Neither Should You): Building Production-Ready Architectures

Darko Mesaros

Are your AI agents exposing sensitive customer data? Stop trusting unpredictable models blindly. Build resilient, production-ready architectures with strict guardrails to contain the blast radius.

Pause
Mute Enter Fullscreen
#1 about 4 min

Recognizing creative failures and risks in AI agents

Real-world agent failures highlight the need to build robust safety systems around large language models.

#2 about 1 min

Categorizing common failure modes in AI agents

Agents typically fail by providing incorrect information, executing harmful actions, or succumbing to prompt injection attacks.

#3 about 2 min

Defining the fundamental AI agent loop architecture

A standard agent architecture combines a large language model, an MCP server for tools, and specific routing SDKs.

#4 about 1 min

Adopting a layered approach to agent safety systems

Designing secure AI applications requires multiple defense dimensions rather than relying solely on the language model.

#5 about 2 min

Implementing guardrails to prevent user abuse and injection

Applying specific guardrails automatically redacts sensitive data and restricts unsafe topics from reaching the underlying model.

#6 about 3 min

Enforcing reliable responses using a steering agent

An intermediate steering language model intercepts and reviews agent responses to catch hallucinations or forgotten tool calls.

#7 about 4 min

Securing data access through deterministic MCP interceptors

Intercepting tool calls to deterministically inject customer authentication prevents agents from hallucinating parameters and leaking cross-account data.

#8 about 6 min

Controlling high-stakes transactions with deterministic policy languages

Evaluating every tool call against defined rules using policy languages prevents agents from executing unauthorized or out-of-order actions.

#9 about 4 min

Establishing robust observability and batch evaluation loops

Implementing continuous tracing and periodic evaluations against chat logs ensures accountability and measures ongoing agent performance.

#10 about 3 min

Combining probabilistic and deterministic controls for AI trust

Trust is established by building layered safety pipelines around the agent rather than inherently trusting the underlying model.

#11 about 6 min

Managing general purpose compute and sub-agent architecture

Limiting agent scope through specialized sub-agents mitigates context limits and improves overall system security and observability.

Matching moments

1:47 min

Why tool-using AI agents require behavioral governance

Sachin Gupta Sachin Gupta · World Congress 2026 North America

2:43 min

Best practices for implementing reliable AI agent frameworks

Marcel Scherenberg Marcel Scherenberg · World Congress 2025

3:34 min

Architecting autonomous agents for production lifecycle management

Sebastian Kister Sebastian Kister · World Congress 2026 Europe

7:41 min

Shattered security assumptions in the age of AI

Javier Garza Javier Garza +4 · World Congress 2026 North America

1:09 min

Resolving developer challenges in AI agent implementation

Ricardo Ricardo · World Congress 2025

4:15 min

Security integration and AI skepticism in developer tooling

Chris Heilmann Chris Heilmann +2 · LIVE

More from World Congress 2026 North America

2026 in LLMs (so far) thumbnail
31:31
10 days ago

2026 in LLMs (so far)

Simon Willison

Running Operations Without Humans in the Loop thumbnail
32:31
10 days ago

Running Operations Without Humans in the Loop

Edward Aryee, Ryan Ries, Yaron Schneider, Chris Hallenbeck, Floriane Gramlich

Beyond the Code: Human-AI Synergies in Product Development thumbnail
17:43
10 days ago

Beyond the Code: Human-AI Synergies in Product Development

Ajita Kanchivakam Ananth

Honey, look! I vibe-coded an OS! thumbnail
30:40
10 days ago

Honey, look! I vibe-coded an OS!

Ian Smith

The Productivity Illusion: What AI Really Changed for Developers thumbnail
24:28
10 days ago

The Productivity Illusion: What AI Really Changed for Developers

Fatih Kadir Akın

Your Evals Passed. Your Agent Just Emptied a Database. thumbnail
21:37
10 days ago

Your Evals Passed. Your Agent Just Emptied a Database.

Tejas Pravinbhai Patel

Related videos

The day the chatbot asked for sudo thumbnail
27:48
World Congress 2026 Europe • 3 months ago

The day the chatbot asked for sudo

Alex Olivier

Responsible AI Architecture with Zero Trust Agents thumbnail
25:33
World Congress 2026 North America • 12 days ago

Responsible AI Architecture with Zero Trust Agents

Ashok Prakash

The AI Agent Path to Prod: Building for Reliability thumbnail
30:48
World Congress 2025 • about 1 year ago

The AI Agent Path to Prod: Building for Reliability

Max Tkacz

Give the Agent a Budget, Not a Token thumbnail
26:00
World Congress 2026 North America • 11 days ago

Give the Agent a Budget, Not a Token

Sachin Malhotra

Securing AI Agent Infrastructure: Identity, Attestation, and Trust at Scale thumbnail
25:54
World Congress 2026 North America • 11 days ago

Securing AI Agent Infrastructure: Identity, Attestation, and Trust at Scale

Abdel Fane

Building Agents Securely at Scale - Alfonso Graziano thumbnail
31:49
Coffee With Developers • 5 months ago

Building Agents Securely at Scale - Alfonso Graziano

Alfonso Graziano

Related articles

From learning to earning

Jobs that call for the skills explored in this talk.