The world of enterprise artificial intelligence is experiencing a surge in both innovation and urgency as companies scramble to keep their rapidly multiplying AI agents under control. In the United States, United Kingdom, and now South Korea, security, governance, and operational oversight have become front and center issues—prompting major technology firms to roll out new solutions that promise to rein in the chaos before it spirals any further.
On April 9, 2026, API management company Gravitee announced the launch of its AI Gateway, a platform designed to control, secure, and monitor the operation of enterprise AI agents. This move comes on the heels of a revealing new report, State of AI Agent Security 2026, which painted a sobering picture of the current state of affairs: US and UK large corporations are now running approximately 3 million AI agents, a number that actually surpasses the global workforce of Walmart, according to the report cited by Gravitee. Even more alarming, 47% of these agents are functioning without any active monitoring or security measures in place. That’s nearly half of all enterprise AI agents operating in the digital wild, unsupervised.
Security incidents are not hypothetical. The report found that a staggering 88% of companies had either experienced or suspected an AI agent-related security or data privacy incident in the past year. As AI agents become more autonomous—accessing internal data, executing external tools, and communicating with each other—the risks to sensitive business information and compliance multiply. It’s no wonder, then, that the business world is shifting its focus from simply creating AI agents to mastering their control.
Gravitee’s AI Gateway aims to tackle this challenge head-on. The platform brings together two of the most critical pathways in AI agent operations: access to large language models (LLMs) and the connection to external tools. Rather than managing these channels separately, the AI Gateway consolidates communication, tool calls, and LLM requests into a single, unified control point. The architecture includes an LLM Proxy and an MCP (Model Context Protocol) Proxy, along with the previously introduced A2A (Agent-to-Agent) Proxy, creating a comprehensive management layer for agent-to-agent communication, tool discovery and execution, and LLM interactions.
What does this mean in practice? For starters, the LLM Proxy allows enterprises to connect with multiple LLM providers—such as OpenAI and Gemini Bedrock—through a single interface, making it easier to enforce policies and monitor activity. This proxy supports token limits (to restrict the volume of data processed), guardrails that block abusive or harmful content, and policies to protect personally identifiable information (PII). It also offers real-time monitoring for latency, performance, and cost, giving organizations visibility into every request and response.
The MCP Proxy, meanwhile, standardizes how AI agents discover and use external tools. It introduces verification, governance, and access controls at every point where an agent interacts with an outside system, aiming to prevent unauthorized or risky behaviors. According to Gravitee, this setup represents an evolution beyond the traditional agent mesh, moving from simple connectivity to a more nuanced approach that also bundles in policy enforcement, security controls, and observability. The goal? To reduce operational complexity as the number of agents continues to grow.
Rory Blundell, Gravitee’s co-founder and CEO, underscored this new direction in enterprise AI: “The focus for enterprises has shifted from ‘can we build agents?’ to ‘how do we control them?’” he said, as reported in Gravitee’s official materials. “AI Gateway provides the essential control layer for organizations to operate AI agents safely.” In essence, Gravitee is positioning AI agents not as experimental tools, but as managed infrastructure—on par with APIs and event-driven systems. The AI Gateway functionality, the company notes, is available for immediate adoption.
But Gravitee isn’t the only player making waves in this space. On April 13, 2026, HS Hyosung Information Systems announced it is considering supplying Hitachi Vantara’s integrated AI agent management platform, Hitachi iQ Studio, to the South Korean market. Originally unveiled in December 2025 and showcased at the NVIDIA GTC conference in March 2026, Hitachi iQ Studio is designed to provide end-to-end management for the full lifecycle of AI agents—from creation and evaluation to fine-tuning and operational governance.
Hitachi iQ Studio is particularly notable for its support of Sovereign AI, a model that allows enterprises to maintain strict control over their data and AI models while building, deploying, and operating agents securely within on-premises environments. This is a crucial feature for industries such as finance, manufacturing, and the public sector, where regulations are tight and data cannot be allowed to stray outside corporate walls.
The platform boasts a range of features tailored to enterprise needs: configurable guardrails to ensure responsible AI behavior, sandbox testing environments for safe experimentation, audit logs for traceability, and conversation history management for compliance. It’s built on reference architecture from NVIDIA’s AI platform, which means it comes pre-integrated for rapid deployment. No-code and low-code agent builders, along with secure Retrieval-Augmented Generation (RAG) pipelines, make it easier for organizations to create and operate AI agents without requiring deep technical expertise.
Hitachi iQ Studio also leverages the Model Context Protocol (MCP) to simplify data access and automation, using predefined templates to streamline the design and deployment of AI agents. This approach is intended to address what HS Hyosung Information Systems, citing a Boston Consulting Group report, identifies as major stumbling blocks for many companies: a lack of AI expertise, inadequate business processes, and limitations in data governance that prevent AI from delivering tangible business value.
By focusing on security, compliance, and cost efficiency, Hitachi iQ Studio aims to enable enterprise-grade AI environments that are both powerful and manageable. HS Hyosung Information Systems emphasized that the platform’s capabilities are especially relevant for organizations facing strict regulatory requirements and those seeking to avoid the pitfalls of data leakage or uncontrolled agent behavior.
As the number of enterprise AI agents accelerates worldwide, the stakes for effective control and security have never been higher. Solutions like Gravitee’s AI Gateway and Hitachi iQ Studio represent a new wave of infrastructure designed to keep pace with the explosive growth of AI—offering businesses the tools they need to harness the power of autonomous agents without losing sight of safety, compliance, or operational clarity.
With these advancements, the enterprise AI landscape is being reshaped, shifting from unchecked experimentation to a new era of managed, secure, and accountable deployment. The race is on—not just to build smarter agents, but to ensure they remain trustworthy partners in the digital age.