AI security’s ‘Great Wall’ problem

The report contends that AI security failures stem less from cloud infrastructure and more from systemic weaknesses such as human factors, supply chains, agent/tool permissions, and change controls. Using the Great Wall analogy and industry statistics, it argues defenders should expand threat models beyond hosting to include data pipelines, plug-ins, agents, and non-human identities, implement zero standing privileges and audit-grade change management, and invest in detection that traces inputs to outcomes. Cloud security is necessary but insufficient; resilient AI systems require governance of delegated authority and robust provenance and observability across the AI ecosystem.