Netskope rolls out AI Guardrails as enterprise AI security demand soars

Netskope said it has made AI Security Guardrails generally available while simultaneously adding discovery and controls for Model Context Protocol traffic, moves the company frames as a response to rapid enterprise adoption of generative AI.

The cloud security vendor, which went public in 2025, markets AI Security Guardrails as a real-time layer that monitors and controls employee interactions with generative AI systems. Company documentation and a LinkedIn post tied to the R133 release describe protections that include defenses against prompt injection and jailbreaking, content moderation of both prompts and model responses, and detection of harmful, biased or copyrighted material. The LinkedIn post stated, "AI Security Guardrails is now available in beta and it brings real time protection for Gen AI usage across the organization," creating an unresolved contrast with other company messaging that calls the feature generally available.

Netskope’s technical materials show how the feature is applied: administrators add AI Security Guardrails profiles to Real-time Protection policies and choose a profile action: Alert, Allow, Block or User Alert. The platform supports API-configurable scanning limits and fallback actions, does not send email notifications for guardrail events, and surfaces matched alerts and events in its Skope IT console. The guardrails map detections to MITRE ATLAS and OWASP Top10 LLM frameworks, giving security teams standardized context for model-related risks.

At the same time, Netskope is extending Netskope One to discover MCP servers and MCP-enabled clients and to collect per-component metadata such as host details, identity attributes and version information. John Martin, chief product officer at Netskope, said, "Every team wants to confidently accelerate AI adoption and emerging protocols such as MCP are now fundamental to that discussion," and warned that "MCP also creates new security risks that legacy tools can’t solve. That’s why we’re further extending the market-leading capabilities of Netskope One to enable teams to see and create policies for MCP traffic and immediately assess how risky MCP tools are."

Netskope frames the product launches against a backdrop of heavy GenAI use inside enterprises. Its Cloud and Threat Report, covering June 1, 2023 through June 30, 2024, shows ChatGPT in use at 80 percent of tracked organizations and Microsoft Copilot at 57 percent, with nearly 200 different generative AI apps monitored. Those adoption figures help explain why vendors are pushing automated controls: agentic AI and protocols like MCP let models connect directly to company data and workflows, expanding the attack surface and creating vectors that legacy security stacks do not see.

The company also highlighted an integration push with key partners and a partner-centric go-to-market strategy, though specific commercial terms and licensing of the guardrails and MCP controls were not disclosed in the available materials. Netskope says MCP discovery and policy controls are available to customers in preview now, with a plan to make the MCP capabilities generally available in the first half of 2026.

For security teams weighing tools, the trade-off is clear: as generative models become embedded in day-to-day workflows, visibility and real-time enforcement move from optional to essential. Netskope’s guardrails and MCP controls aim to turn that visibility into policyable actions, but discrepancies in published availability status imply that customers should check which features are in beta, preview or general release before deployment.