Virtue AI pulls the rug out from under the feet of shadow AI agents

Artificial intelligence security startup Virtue AI Inc. is expanding its capabilities with the launch of Shadow AI, a new addition to its AgentSuite platform that gives enterprises clear visibility into how AI is being used across their organization.

Enterprises desperately need to know which of their employees are using AI, and how they’re using it, for the technology is becoming more prevalent than ever, the company said. Autonomous AI agents in particular are spreading like wildfire inside users’ laptops, software-as-a-service applications, in DevOps workflows and other kinds of platforms. In most cases, organization’s security teams have never had a chance to review, let alone approve, all of these AI systems.

That’s not to say they don’t know about it. Virtue AI says most businesses are only too aware of the rate of AI’s adoption among their workforce, and in many cases they encourage it. But it’s also very risky, because security teams do not have the ability to see how AI is operating within the organization. They don’t know exactly where those AI agents are running, what permissions they’ve been given or what tools they can use. They don’t have a way to track their behavior, and they certainly cannot tell if they’re compliant.

Virtue AI is well-qualified to do something about this mess. It’s the creator of an enterprise-grade AI security and compliance platform, with a focus on continuous security testing and real-time safety guardrails. Its systems are designed to protect large language models and AI agents from cyberattacks, including prompt injection and jailbreaking, and it can also prevent agents leaking confidential data.

Its expertise means it’s well-placed to understand what AI agents are doing within different organizations, and it also knows the risks of having them run around doing whatever they want with no restrictions. AI agents could, for example, accidentally access and delete data they were never intended to touch.

Shadow AI is the company’s response. It’s an all-new endpoint-level discovery and monitoring layer that’s built especially to detect and track AI agents. According to Virtue AI, it works very differently from traditional endpoint detection and response and extended detection and response tools, which treat AI agents as if they were any other application. It’s designed to identify where agents are running, understand how they plan, act and use tools, and observe how they evolve over time.

“Across the enterprise, employees are using unapproved agents for things like coding, data analysis and sales outreach,” said Head of Agent Security Wenbo Guo. “We built Shadow AI to find them.”

Perhaps the most important capability of Shadow AI is its ability to find AI agents buried within every application, data pipeline, browser extension and more. Once it discovers an AI agent, it will start monitoring its activity, especially things such as its process activity, its network behavior and any file system changes it makes.

Shadow AI allows security teams to capture the full behavioral sequence of each AI agent, so they can catalog the order, structure and context of their actions. They’ll be able to map this activity across their entire information technology environment, in order to see at a glance how many agents are active, on what devices and what policies they may violate. Teams will then be able to distinguish an agent’s normal behavior, and flag it when it starts acting suspiciously.

Should an agent start misbehaving, teams will have a full record of its host, user context, tool calls and action sequence, so they can quickly work out what has happened. If necessary, they’ll be able to use Shadow AI to take the agent offline, and then work out how to fix any damage it has caused.

Shadow AI can be deployed as a light endpoint collector. It’s compatible with Linux, Windows and macOS systems, and can be used either independently or alongside existing EDR and XDR platforms, including CrowdStrike Falcon and Microsoft Defender. It can also be used to aggregate data sources into a single pipeline to simplify agent pattern detection, so teams can work with a single, unified view across their entire endpoint and SaaS signals.

Virtue AI says it will provide immediate value to any enterprise that’s leaning on AI agents. With full visibility, they’ll be able to keep tabs on anything they do, immediately flag anything suspicious, carry out a comprehensive and rapid investigation, and eliminate much of the risk that comes with them.

“Shadow AI surfaces the agents running in your environment, traces their actions and shows your team what each agent is doing, so you can confidently scale AI across your business,” Guo said.

Image: SiliconANGLE/Gemini