Box Inc. (NYSE: BOX), a prominent player in the realm of Intelligent Content Management (ICM), has made a significant stride in enhancing the security of enterprise content involving artificial intelligence (AI) agents. In a press release dated July 21, 2026, the company unveiled new security capabilities that aim to give organizations better control over how AI agents access and utilize sensitive data. This initiative comes at a time when a substantial 83% of organizations are actively experimenting with AI agents in their operations.

2. Addressing Security Challenges in AI Deployment

As organizations increasingly integrate AI into their workflows, security and privacy have emerged as critical concerns. A recent report by Box highlighted that 90% of IT leaders view security, regulatory, and trust issues as the main barriers to deploying AI agents at scale. Manoj Asnani, VP of AI Security, Privacy, Compliance & Governance Products at Box, emphasized the necessity for robust security measures that allow AI agents to function effectively without compromising sensitive information.

"As these agentic workflows become more deeply embedded in the enterprise, it’s critical to create the proper security controls," Asnani stated. The newly introduced features aim to mitigate risks associated with AI deployment, ensuring that organizations can confidently scale their use of AI agents.

3. New Security Features

Box’s new security capabilities are integrated directly into the platform, allowing organizations to apply stringent security measures at every point of interaction with their content. The key features include:

Agent Guardrails: These establish strict parameters for what custom Box AI agents can do, based on content sensitivity and organizational policies. This includes enforcing access controls and requiring approval for certain actions.

Prompt Injection Detection: This feature validates every input to detect and neutralize known prompt injection patterns, thereby preventing unauthorized access or modifications to data.

MCP Guardrails: Admins can control the actions of external AI agents connected via the Box MCP Server, ensuring that these agents operate within approved boundaries.

Classification-Based Access Policies: Organizations can set specific classifications to restrict access to sensitive content, ensuring only authorized agents can interact with classified data.

Agent Activity Oversight: This provides visibility into the activities of external AI agents, enabling organizations to set alerts for suspicious behavior and respond quickly.

Agent Audit Trails and Session Governance: Every agent interaction is documented for compliance and audit purposes, making it easier for organizations to maintain oversight.

Human-in-the-Loop Control: This feature mandates human approval for sensitive actions performed by AI agents, adding an extra layer of security.

These capabilities are designed to facilitate the transition from pilot projects to large-scale AI deployments, all while maintaining rigorous governance standards.

4. Industry Applications

The enhancements in security and governance are particularly relevant for several sectors:

Financial Services: Firms can safeguard sensitive information related to mergers and acquisitions and trading activities by enforcing agent-level access controls and monitoring agent behavior in real-time.

Healthcare: Organizations can protect patient data and proprietary research through classification-based access controls, ensuring compliance and security.

Legal: Law firms can govern AI's role in contract analysis and discovery workflows, maintaining control over what information agents can access.

Insurance: Companies can validate inputs from AI agents to prevent unauthorized access to sensitive policyholder data, ensuring regulatory compliance.

5. Expert Insights

Industry experts have lauded Box’s initiative as a timely response to the growing need for secure AI deployment. Tatsutoshi Murata, Head of IT Strategy Department at Nomura Research Institute, expressed confidence in Box's ability to provide the necessary administrative features to navigate this new era of AI. “With a protective layer that appropriately manages AI agent access to content, we’re confident our critical content will remain protected as we expand our use of AI,” Murata remarked.

Amy Machado, Senior Research Director at IDC, echoed this sentiment, stating that Box's new security controls directly address the primary barriers of privacy and unauthorized access.

6. Conclusion

Box’s latest security enhancements represent a significant advancement for organizations looking to leverage AI agents safely and effectively. With these new capabilities, Box continues to set a standard for security in the AI space, reinforcing its commitment to protecting sensitive information while facilitating collaboration and efficiency.

7. Availability

The new security and governance capabilities will be available to customers on the E-Advanced plan in the coming months. For more information, visit the Box Blog.

Original press release