Box Enhances AI Governance with New Security Features for Enterprise Content

• •Box introduced new security features for AI agents managing enterprise content.
• •New capabilities include oversight, prompt injection detection, and access policies.
• •The enhancements aim to support organizations transitioning to full production use.

Box has launched new security capabilities aimed at enhancing governance over AI agents interacting with enterprise content. These features include agent guardrails, oversight of third-party agent activities, prompt injection detection, and access policies based on agent classification. The new controls allow organizations to monitor, record, and restrict agent actions at the content layer, ensuring that they remain within approved permissions. This initiative is designed to facilitate the transition from pilot programs to full production use without sacrificing visibility or control. Affected parties include organizations utilizing Box's services and third-party AI models like Claude, ChatGPT, and Gemini. The enhancements are part of Box's broader effort to extend its security framework to agentic workflows, promoting safe deployment of AI technologies.