Box Introduces AI Agent Security Framework for Enterprise Content — Box, Inc.

Event summary

Box launched new security controls for AI agents operating on enterprise content, including agent guardrails, prompt injection detection, and classification-based access policies.
The capabilities extend to both Box Agents and third-party agents like Claude, ChatGPT, and Gemini.
83% of organizations are already experimenting with AI agents across critical tasks, per Box's VP of AI Security.
Box’s 2026 State of Enterprise AI report found that 90% of IT leaders cite security as the biggest barrier to deploying AI agents at scale.

The big picture

Box's move addresses a critical gap in enterprise AI adoption: securing content accessed by AI agents. As organizations increasingly rely on agentic workflows, governance frameworks that balance functionality with oversight will become table stakes. The introduction of these controls positions Box as a leader in AI-driven content management, particularly for industries handling sensitive data.

What we're watching

Adoption Pace: How quickly enterprises will integrate these security controls into their AI agent workflows, given the current 83% experimentation rate.
Competitive Response: Whether competitors in content management and AI governance will introduce similar frameworks to secure enterprise content.
Regulatory Alignment: The extent to which Box’s new controls align with emerging regulations on AI and data security across industries like financial services and healthcare.