Reco Aims to Eradicate Shadow AI Blind Spots Across Enterprises

AI is spreading through workplaces at an unprecedented pace. Every day, employees connect AI tools to enterprise systems—often without approval or oversight from IT security teams. The result is what experts call shadow AI: a growing network of integrations and tools that access company data without monitoring.

Dr. Tal Shapira, Co-founder and CTO at Reco, a SaaS security and AI governance provider, says this invisible sprawl could become one of today's biggest organizational threats. He notes that the speed of AI adoption has already surpassed current enterprise safeguards.

“We moved from ‘AI is coming’ to ‘AI is everywhere’ in roughly 18 months. The challenge is that governance frameworks haven’t kept up,” Shapira said.

The hidden threat within corporate systems

According to Shapira, most corporate security systems were built for an older era, where everything stayed behind firewalls and network boundaries. Shadow AI disrupts that model by operating from within, concealed inside the company's own applications.

Many modern AI tools integrate directly with everyday SaaS platforms like Salesforce, Slack, or Google Workspace. Though this isn’t inherently dangerous, AI often connects through permissions or plug-ins that remain active after setup. These “silent” connections may continue providing AI with access to company data—even after the employee who installed them leaves or stops using the tools. This represents a major shadow AI challenge.

Shapira said: “The deeper issue is that these tools embed themselves into company infrastructure, sometimes for months or years without being detected.”

This new class of risk is especially hard to track since many AI systems are probabilistic. Rather than executing fixed commands, AI makes predictions based on patterns. As a result, their behavior varies by context, making oversight and control more difficult.

When AI behaves unexpectedly

Real-world incidents reveal the tangible impact of shadow AI. Reco recently collaborated with a Fortune 100 financial firm that believed its systems were secure and compliant. Within days of deploying Reco's monitoring, the company uncovered over 1,000 unauthorized third-party integrations across its Salesforce and Microsoft 365 environments—more than half driven by AI.

One integration, a transcription tool linked to Zoom, had recorded every customer call—including pricing negotiations and confidential feedback. “They were inadvertently training a third-party model on their most sensitive data,” Shapira explained. “There was no contract nor clarity on how that data was stored or used.”

In another case, an employee connected ChatGPT directly to Salesforce, allowing the AI to produce hundreds of internal reports in hours. While efficient, this exposed customer details and sales forecasts to an external AI platform.

How Reco reveals what goes unseen

Reco's platform provides companies with full visibility into which AI tools connect to their systems and the data those tools can reach. It continuously scans SaaS environments for OAuth permissions, third-party apps, and browser extensions. Once identified, Reco indicates which users installed them, what permissions they have, and whether the behavior looks unusual.

If a connection appears risky, the system alerts administrators or automatically revokes access. “Speed is critical because AI tools can extract huge data volumes in hours, not days,” Shapira noted.

Unlike traditional security products focused on network perimeters, Reco targets the identity and access layer. This makes it well-suited for today’s cloud-first, SaaS-dependent organizations, where most data resides beyond the traditional firewall.

A broader security alert

Industry analysts say Reco’s approach reflects a larger enterprise security trend: shifting from blocking AI to governing it. A recent Cisco report on AI readiness found that 62% of organizations admit having limited visibility into how employees use AI tools at work. Nearly half have already faced at least one AI-related data incident.

As AI becomes embedded in mainstream software—from Salesforce’s Einstein to Microsoft Copilot—the challenge escalates. “You might assume you’re using a trusted platform,” Shapira stated, “but not realize it now includes AI features that automatically access your information.”

Reco’s platform addresses this gap by monitoring both approved and unapproved AI activity, helping organizations understand where their data flows and why.

Adopting AI with confidence

Shapira believes we're entering the “AI infrastructure phase”—a time when every business tool will include AI in some form, visible or not. That makes continuous monitoring, least-privilege access, and time-limited permissions essential.

“Successful companies won’t be those that block AI,” he observed. “They’ll be the ones that adopt it responsibly, with guardrails that protect both innovation and trust.”

Shadow AI, he emphasized, doesn’t reflect employee carelessness, but how fast the tech is advancing. “People want to be productive,” he said. “Our role is to enable that without placing the organization at risk.”

For enterprises seeking to leverage AI without losing control of their data, Reco’s message is straightforward: You can't secure what you can't see.

Image source: Unsplash