Box unveils new controls to secure AI agents operating across enterprise content

Box announced planned security and governance capabilities for AI agents that interact with enterprise content, extending its controls to both custom Box Agents and third-party tools such as Claude, ChatGPT, and Gemini. The features include policy-based agent guardrails, classification-based access restrictions, prompt-injection detection, controls for agents connected through the Box MCP Server, activity monitoring and alerts, audit trails with retention and legal-hold support, and human approval requirements for sensitive actions. Box says the controls operate at the content layer to help organizations govern agent access, actions, and data use across industries handling sensitive information, including financial services, healthcare, legal, and insurance. Citing its survey finding that security, regulation, and trust concerns impede large-scale AI-agent adoption, the company positions the release as an expansion of its existing Box Shield security offerings. The capabilities are expected to roll out to customers on the E-Advanced plan in the coming months, subject to the product-development and market risks described in the release.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.