Lakshmi Harika Akkireddy
Lakshmi Harika Akkireddy
Securing the Sky: Why Cloud Security Breaks Down When AI Moves In
Cloud security was never simple, but for years we at least understood the rules. You protect the perimeter, manage access, encrypt the data, monitor the logs, and respond when something goes wrong. It was hard work, but it was knowable work. Then AI moved in, and the rules stopped making sense. AI workloads do not behave like traditional applications. They pull in massive volumes of sensitive data for training. They spin up compute resources dynamically and at a scale that makes manual oversight almost impossible. They communicate with external APIs, third-party model providers, and data pipelines in ways that existing cloud monitoring tools were never built to track. And because they learn and adapt over time, a system that was secure when you deployed it may not stay that way six months later without anyone touching the configuration. This session is an honest look at where the traditional cloud security playbook breaks down when AI enters the picture, and what it takes to build something better. We will start with the structural cracks: why shared responsibility models create accountability gaps the moment AI workloads are involved, why identity and access management frameworks designed for human users fail when the entity requesting permissions is an autonomous system, and why the data governance practices most organizations have in place were simply not built for the volume and velocity AI demands. From there, we move into what is genuinely new. Not just variations on old threats, but emerging attack patterns specific to AI in the cloud, including model poisoning, inference attacks, training data extraction, and the exploitation of overprivileged AI agents operating inside cloud environments with little to no oversight. The session closes with a practical roadmap. What security teams need to change immediately, what investments are worth making now versus later, and how to have the conversation with leadership before a breach makes it unavoidable. Cloud computing changed everything about how we build software. AI is changing everything about how cloud computing works. The security strategies that got us here are not the ones that will keep us safe going forward.