OpenClaw went from a weekend project to 200,000+ GitHub stars in weeks, and with it came developers spinning up cloud VMs to run autonomous AI agents 24/7. But cloud infrastructure was built for stateless, request-driven workloads. Agents are none of those things. They hold state, wake up on timers unprompted, inherit broad system permissions, and chain together API calls, file operations, and browser sessions in ways no one planned for.
This talk covers what breaks when agents hit cloud infrastructure. First, the architecture mismatch: why serverless and container platforms struggle with stateful, long-running, proactive workloads. Then the security surface, including supply chain attacks through unvetted skill repos, prompt injection as an infrastructure-level threat, and the gap in cloud IAM for non-human agent identities, grounded in real incidents and the OWASP Top 10 for Agentic Applications (2026).
Finally, what works. Reliability patterns from SRE: least-privilege scoping, container isolation, observability for agent workflows, and error budgets for destructive actions. You'll leave with a practical checklist for deploying agents on cloud safely.