SREs are the guardians of reliability — we build for failover, redundancy, and scale. But in today’s AI-native systems, there’s a quiet shift happening: automation scripts, AI-driven agents, and machine identities are running critical operations with increasing autonomy.

They auto-remediate incidents, make scaling decisions, deploy configurations, and even patch systems — often with elevated permissions and little human oversight.

But here’s the problem:

• What happens when these agents act on bad data?
• Or when a misconfiguration triggers a cascade of failures?
• Who holds accountability when an over-permissioned automation silently breaks production?

In this session, I will expose the blind spot most SRE teams don’t realize they have — the growing influence of non-human actors in reliability engineering. I will share real-world examples of automation gone wrong, highlight how over-trusted machine identities quietly amplify risk, and explain why SREs need a governance mindset toward automation.

I will discuss:

• The evolving role of AI agents and automation in SRE workflows
• How machine identities can quietly become single points of failure
• Why over-permissioned bots and lack of observability threaten reliability
• Practical steps for introducing identity governance, privilege boundaries, and automation observability into SRE practices
• How to chaos-engineer for AI agents and validate fail-safes

This is not a security talk — it’s a reality check for SREs operating in environments where automation does more than assist; it decides. If you are an SRE, platform engineer, or operations lead, this talk will challenge you to rethink your approach before your AI helpers become your biggest liability.

Cynthia Akiotu is a Cybersecurity Architect and Identity Specialist with deep expertise in Identity & Access Management (IAM), data governance, and Zero Trust implementation. She focuses on securing environments where AI agents, autonomous models, and machine identities are rapidly expanding the attack surface. With experience across diverse sectors, Cynthia has led projects on secure cloud adoption, privileged access management, insider risk mitigation, and identity governance for AI-driven environments. She holds several industry certifications, including Microsoft Certified: Identity & Access Administrator and Microsoft Certified: Information Security Administrator. Her thought leadership extends to academic and industry publications, contributing to works on AI-enabled regulatory compliance and digital skills development in security. Beyond her professional work, she volunteers as Cyberhero Coach Cynthia, promoting cybersecurity awareness and safe digital habits among children and communities. Driven by a commitment to secure innovation, she champions identity governance as a foundation for trust in AI-native ecosystems.