When approaching AI security, federal agencies cannot treat the technology simply as another application deployed behind the network perimeter.

Once an AI system can retrieve sensitive information, communicate with other software and act through application programming interfaces (APIs), it becomes an operational participant with its own access, dependencies and potential attack paths. Agencies are experiencing this transition while adversaries begin deploying their own AI-powered tactics to accelerate reconnaissance and attacks.

In this era, securing AI must go beyond protecting the underlying model to include governance of every identity, prompt, data source, tool connection and action surrounding it.

DISCOVER: Work more securely with built-in AI across every layer of technology using Google devices.