Surprisingly Useful AI Article Enhancements
Netskope has expanded its partnership with Google Cloud to deliver a security solution for generative AI and autonomous agentic AI workflows running on Google Cloud infrastructure.
The solution, called Netskope One AI Guardrails, is designed to apply real-time safety checks to AI workloads – including autonomous agents that execute tasks independently – without introducing significant latency. It runs on Google Cloud’s Tensor Processing Units (TPUs) and uses the Vertex AI platform for real-time content moderation.
The agentic AI security problem
Enterprise investment in AI is projected to exceed US$867 billion worldwide by 2029. As organisations move beyond simple chatbot deployments toward autonomous AI agents that can call tools, interact with external services and execute tasks without human intervention, the security implications become more complex.
Traditional security approaches that focus on data at rest are not designed for the dynamic, multi-step interactions that agentic AI systems generate. An autonomous agent might call a tool, query an API, interact with a Model Context Protocol (MCP) server and execute a task in sequence – each step representing a potential attack surface.
Netskope’s solution is designed to verify each of those steps against enterprise policy in real time, inspecting the data passed between agents and tools to prevent threats like indirect prompt injection and malicious execution.
Hardware-accelerated security
A key technical claim in the announcement is around latency. Security checks on AI workloads can introduce delays that undermine the performance benefits of running models on accelerated hardware.
By running its guardrails on Google Cloud TPUs – the same specialised accelerator architecture that supports Google Cloud’s own AI services – Netskope argues it can perform safety checks in line with the model’s native processing speed, maintaining low latency without compromising security coverage.
The approach uses Google Cloud’s Vertex AI platform to perform real-time moderation, aligning the security layer with the infrastructure already underpinning the AI workloads it is protecting.
What the guardrails cover
The solution enforces real-time safety checks across several categories, including hate speech and discrimination, criminal activity, suicide and self-harm, piracy and copyright and requests for sensitive data.
On the threat detection side, Netskope targets AI-specific attack vectors including prompt injection and jailbreaking attempts. It also detects and blocks unintended recursive loops or what the company calls “ghost” commands – malicious instructions that can lead to resource exhaustion or unauthorised system modifications.
All detections are mapped to industry-standard frameworks including MITRE ATLAS and the OWASP Top 10 for large language models, providing security teams with an auditable risk posture.
Data sovereignty
The solution is deployed directly within a customer’s Google Cloud instance, meaning that prompts and responses are scanned locally and data does not leave the governed tenant.
That architecture is designed to help customers meet compliance requirements under frameworks like GDPR, HIPAA and the EU AI Act by enforcing consistent policy across all AI communications within the customer’s own environment.
For organisations in regulated industries or jurisdictions with strict data residency requirements, this deployment model addresses a common concern about sending sensitive data to external security services for inspection.
Partner perspectives
Vineet Bhan, Director of Security and Identity Partnerships at Google, framed the collaboration around combining AI capability with enterprise security.
Related: Best Business Laptops for work & school
Related: Best Gaming Laptops
Related: Best Portable Laptop
“Netskope One AI Guardrails works with Vertex AI and TPUs to enable the secure deployment of both generative AI and autonomous agents,” Bhan explained. “Together, we are helping organisations build the future of their business on a foundation of trust.”
Sanjay Beri, Chief Executive Officer and Co-Founder of Netskope, pointed to the importance of security keeping pace with AI adoption.
“Netskope provides critical capabilities required for the safe, high-performance adoption of modern AI workflows,” Beri noted. “We are proud to collaborate with Google Cloud to help ensure that customers can confidently pursue their AI goals.”
Last Updated on April 23, 2026 by Nick Ross
