Talk
Virtual
How to productize GenAI guardrails as a platform service
In this talk, the speaker shows how platform teams can offer guardrails as a shared service so AI-enabled apps receive prompt injection, data leakage, and policy checks, with observability and continuous testing, without slowing teams down.
CEST
Meet the speakers
The speaker shows how a platform team can productize guardrails as a service to help ensure GenAI features ship safely. Firewalls and authentication do not stop prompt injection, output data leaks, or policy violations, and application teams should not have to reinvent the same controls.
They present a modular approach, using a proxy or orchestrator with pluggable detectors, that lets teams mix controls by use case, including prompt injection, content safety, PII and secrets, language, and custom business rules, while the platform mandates baselines. The speaker also illustrates the approach with a live demo that blocks attacks and publishes metrics.