Fortinet pairs FortiAIGate with OCI Enterprise AI

Fortinet's post frames the shift from isolated AI experiments to production systems as the reason security can no longer be an afterthought. The stated risk goes beyond prompts and responses to the actions agents take — a Wall Street Journal report cited in the post describes an agent sharing personal financial information in a company Slack channel, showing how quickly a routine task can become an incident. Without a unified governance framework, the post argues, teams apply different permissions and data-access policies, creating gaps that attackers or unintended AI behavior can exploit.

The proposed split of duties is between Oracle's managed layer and Fortinet's runtime inspection. OCI Enterprise AI supplies managed model access, agent tooling and platform governance; FortiAIGate sits between applications and models as a single inspection and control point, and in the reference architecture it runs inside the customer's own virtual cloud network on OKE worker nodes, with a hub-and-spoke option via Oracle Dynamic Routing Gateway. Because the gateway is self-hosted, inspection and logs stay in the organization's environment.

FortiAIGate's controls extend to AI spending and tool use: per-workspace budgets that reset on a daily, weekly or monthly basis, rate limits per minute, and MCP tool scanning in both directions to catch prompt injection and data leakage. The post notes that where and how models are processed depends on the chosen provider and deployment configuration, so organizations still need to select routing and models that meet their data-residency requirements.