Sponsored Keynote: Rescue Agents From Prototype Purga...
summary
Defines concrete readiness checks for agents: resilience to model changes, strict data-exfiltration controls, and behavioral boundaries, with an explicit target of <2% hallucination rate as an example threshold. Recommends enforcing policies via network-layer mediators like Envoy proxy and extensions such as Envoy gateway / Envoy AA gateway that inspect traffic, redact PII, and apply fallbacks. Describes an implemented stack—Amboy AI Gateway and Agent Router Enterprise—used to gather metrics, run factual consistency checks, and automate guardrail actions.
AI-written summary · about
related talks
19mKeynote: Towards Building Safe and Secure Agentic AI
40mIn AI We Trust? Securing the Future, One Agent at a Time
30mWhy Prompt-Level Guardrails Fail for AI Agents
18mFrom Reactive to Resilient: Agentic AI
17mGuardrails First: Designing AI Agents for Security Operations
30mArchitecting Integrity: Responsible Software Evolution in the AI Era
26mBeyond Hype: Leading the Charge in Responsible AgentOps for Business Transformation
11mBuilding Secure Backend Services with AI Agents