Netskope launches security controls for autonomous AI agents

Netskope has announced Netskope Skylight Agent Action Control, a new security capability designed to classify and block high-risk actions taken by autonomous AI agents before they are executed. The tool aims to address a significant security gap, as the company reports that 91% of organizations are currently unable to stop risky agent actions before execution, and 54% have experienced confirmed or suspected AI agent security incidents in the past year.

The new capability categorizes agent actions into nine intent-based types, including data destruction, credential manipulation, remote code execution, and potential data exfiltration. Security teams can apply risk-based policy profiles—ranging from low to critical—to specific agent types, allowing for different rules for coding assistants versus chat applications.

As part of this launch, Netskope is rebranding its AI security portfolio from Netskope One AI Security to Netskope Skylight. This suite includes the AI Command Center, AI Guardrails, AI Gateway, Agentic Broker, and AI Red Teaming. The Agent Action Control feature is expected to be available by the end of the third quarter of 2026.