Securing AI with strict boundaries
Once your agents can reach real systems and act on live data, the question stops being "can it do this?" and starts being "should it, and who's accountable if it doesn't stop itself?"
This session digs into the consequences of not properly implementing the right boundaries and guardrails for AI agents. We start with the Moffatt v. Air Canada case — a chatbot that invented a bereavement fare policy that didn't exist — as a leading example. From there, we explore the full landscape of agent risk: hallucinated policy, PII leakage, prompt injection, jailbreak attempts, and the more mundane but still damaging problem of an agent wandering into an off-topic, embarrassing tangent.
We get concrete about the guardrail mechanisms that actually hold these boundaries: topical guardrails that draw a hard line around specific subjects, ethical moderation layered on top of a model provider's defaults, jailbreak detection, PII redaction with support for custom entities, and groundedness—forcing an agent to answer only from the context it's been given.
What You'll Learn:
Why "the model hallucinated" is a legal liability problem wearing a technical disguise, and what Moffatt v. Air Canada actually settled
The full risk surface behind agent guardrails: hallucinated policy, PII leakage, prompt injection, jailbreaks, and off-topic brand risk
How topical guardrails, ethical moderation, jailbreak detection, PII redaction, and groundedness each close a different gap
Why groundedness is the guardrail that would have stopped Air Canada's chatbot cold
How guardrails sit on top of any reasoning pattern (React, Reflection, Deep Agent) as a final, non-negotiable check
Why a SOC 2 or HIPAA badge tells you how a vendor runs their own house, not whether your specific rules are being enforced
A live look at configuring topical rules, PII entities, and groundedness inside the Ejento guardrails panel