Build a Platform and Watch It Burn — Michael Forrester, Accenture & Whitney Lee, Datadog

Demonstrates that refusal behavior is not a boundary: excessive agency and poisoned tool descriptions need infrastructure controls, filtering and tracing to catch.

• In this workshop a deliberately permissive burrito-ordering bot is pushed into deploying an external container image and exposing a secret recipe from its Kubernetes environment, making excessive agency concrete.
• Model refusals varied between attempts, which the presenters use to argue an application's boundaries cannot depend on the model choosing to refuse. Kyverno rules restricting image sources helped, but intended runtime protections did not consistently stop recipe access.
• A weather tool whose description asks the agent to read internal configuration shows tool metadata as a trust boundary. LLM Guard input and output filtering redacted or blocked sensitive content, and OpenTelemetry traces in Datadog made prompts and tool calls inspectable.

• AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
• LLM — A large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.

• videoHow Secrets Leak Through AI Agents (and How to Stop It) — Venice AI
• videoFrom Zero to Leaderboard: Agent Evaluation — Wolfram Ravenwolf, Weights & Biases
• videoContinuously Improving Agents with Langfuse — Annabell Schäfer & Lotte Verheyden, ClickHouse
• videoLet Your Agent Cook: Using Skills to Evaluate and Improve Your App — Ankur Duggal, Arize AI

• videoIT Admin for the AI Workforce — Sarthak Aggarwal, DecaworkAI Engineer
• videoThe Lethal Trifecta Is Already on Your Laptops — Michael Patterson, CoderAI Engineer
• videoThe Dark Arts of Web Automation: Teaching Agents to Use Websites Like Humans — Corey Gallon, RexmoreAI Engineer

Checking sign-in…

Loading comments…