NordLayer Launches AI Scout to Expose Shadow AI and Block Sensitive Prompts

According to NordLayer, AI Scout monitors browser and desktop interactions. It can identify authorized and unauthorized AI services, allowing companies to compare observed activity with software approved for workplace use. Its reporting measures usage from real activity rather than purchased subscriptions, and it estimates token consumption across deployed large language models.

The tool also examines information employees send to AI systems. NordLayer says controls are designed to detect potentially sensitive material before certain prompts are submitted, and browser-based controls can intervene when employees enter information requiring additional scrutiny before transmission. The monitoring extends beyond conventional chatbot use into AI agents and their technical relationships with connected services used at work. Administrators can inspect MCP infrastructure linked to programming agents and review the capabilities available through those systems in practice.

“We built NordLayer AI Scout for how AI actually works — not just what it looks like from the outside, but for what’s happening at a technical level,” said Vakaris Noreika, VP of product research and development at NordLayer. “Its AI DLP is designed to inspect prompts and control them pre-submission, to ensure that sensitive data is never inputted in the first place.”

NordLayer points to growing business investment and employee use of unapproved services. Writer's 2026 enterprise survey found that 59% of companies spend more than $1 million annually on artificial intelligence technology, according to the report. A University of Melbourne report cited by NordLayer found that 42% of employees use approved AI systems, while 70% use free models at work. IBM reports that breaches involving shadow AI add $400,000 to the global average breach cost of $4.99 million worldwide.

“What companies are facing today is an AI zoo — dozens of unsanctioned LLMs in daily use across the workforce, each one widening the organization’s exposure to data leaks and security breaches,” said Andrius Buinovskis, VP of product strategy at NordLayer. “True AI governance requires extensive visibility. With new LLM solutions emerging at a rapid pace, reviewing every new AI tool manually simply doesn’t scale — organizations need telemetry that brings all AI use, sanctioned or not, into one platform.”

The push for AI visibility comes as browsers take on a larger role in work and personal computing. In a separate TechRadar article published on Oct. 14, 2026, Opera's head of security wrote that browsers now hold login credentials, manage permissions, block trackers and increasingly give AI agents the ability to act on a user's behalf. The article argued that three common misconceptions can misdirect caution.

One misconception is that the browser itself tracks everything users do. The article said data collection in most cases comes from websites, advertising technology, embedded trackers and third-party scripts, not the browser itself, and that browsers can be the layer that shows users when a site wants location access, exposes cookie controls and privacy settings, blocks trackers and flags dangerous connections. Another misconception is that an AI assistant inside a browser automatically sees everything. The article said access can be scoped to a specific page, tab or task, and should be treated like employee privileges: an AI drafting an email should not need access to a banking session in another window, and access should end when the task ends. Security teams should ask vendors what default permissions are, whether access is scoped to a task or granted broadly, and whether users are informed about what the AI is doing.

The third misconception is that privacy and security are the same. The article said security is primarily about preventing unauthorized access or stopping malicious activity, while privacy is about controlling what information is collected, shared and retained. Anonymity can improve privacy by making less activity traceable, but taking it too far can work against security by making malicious activity harder to identify.