Enterprises are moving fast on AI governance and AI security controls. In production, however, controls can fail, drift or be bypassed, and AI dependencies can degrade (throttling, provider outages, model drift and pipeline failures).

This session reframes the problem from 'prevent every failure' to 'ensure AI-enabled services fail safely'. You'll get a practical, operations-first blueprint for AI resilience: how to design fallback modes (human-in-the-loop, rule-based degradation, feature kill switches), how to define AI-specific incident triggers and escalation criteria, and how to test readiness through scenario-based exercises that integrate security, SRE and business owners. The goal is simple: keep critical services trustworthy and available, even when the AI layer becomes unreliable. Attendees leave with a concrete checklist to operationalise resilience across the AI lifecycle – without duplicating governance work or creating control theatre.