Netskope Targets AI Agent Security Gap With Skylight Launch

Netskope Inc. (NASDAQ:NTSK) says 91% of enterprises lack the capability to mitigate risky actions by AI agents before those actions execute. The company is betting its new product can close that gap.

The cybersecurity firm has announced Skylight Agent Action Control, a capability that lets security teams dictate AI agent behavior by classifying risky agent actions and controlling them before execution. The stakes are concrete: 54% of organizations using AI agents have experienced a suspected or confirmed security breach in the past 12 months, according to figures cited in the announcement.

Netskope is positioning the tool as a proactive play rather than a reactive one. The company aims to lead the emerging AI agent governance segment, and Skylight Agent Action Control is the centerpiece of that push.

A Rebranding With Intent

The launch coincides with a broader rebranding of the company's AI security portfolio from Netskope One AI Security to Netskope Skylight. The renamed portfolio now spans five products: AI Guardrails, AI Command Center, Agentic Broker, AI Gateway, and AI Red Teaming. The rebranding consolidates Netskope's AI security offerings under a single, more visible banner.

The announcement extends a stretch of product momentum. In its recently reported second quarter, Netskope rolled out the Netskope One DataSec Command Center, a unified platform for tracking and protecting sensitive data across endpoints, cloud, and AI environments. The company also added features to its NewEdge AI Fast Path capability, which company testing showed can reduce latency by as much as 90% to popular AI destinations.

The Risks Skylight Aims to Stop

Netskope identifies several failure modes that Agent Action Control is built to address. The first is "authority drift," in which an AI agent's permission footprint expands beyond what administrators originally approved. Left unchecked, an agent quietly accumulates access it was never granted.

The second risk involves vague or overly permissive instructions. Ambiguous prompts can lead agents to attempt high-risk actions, such as deleting a production repository — a scenario where a single misunderstood instruction can take down critical infrastructure.

The third problem is timing. Many organizations simply cannot intervene in real time. With the majority of enterprises unable to block a risky agent action as it happens, security teams stay reactive, investigating damage after the fact rather than preventing it. These risks, taken together, expose a gap between enterprise oversight and agent autonomy, particularly around unauthorized external communication and data destruction.

The Commercial Question

Agent Action Control fits squarely into a market where buyers are under pressure. A security breach rate of 54% among AI agent adopters gives vendors a sharp sales argument, and Netskope's pre-execution control model addresses the timing problem that most enterprises admit they cannot solve on their own.

But product capability does not guarantee revenue. The payoff from this investment will depend on how quickly enterprises adopt Agent Action Control, and whether Netskope can monetize the broader Skylight portfolio in an AI security market that is becoming increasingly competitive. Established security vendors and well-funded startups are racing to offer similar agent governance tools, and enterprises will weigh Skylight against alternatives as they formalize their AI oversight budgets.

For now, Netskope has moved early with a named product in a category where the demand signals — a 91% capability gap and a 54% breach rate — are unusually specific. The next test is conversion: turning an acknowledged industry gap into contracted Skylight deployments.