Operant AI finds Shadow Escape, an MCP vulnerability exposing hidden data leaks across leading AI platforms.

Shadow Escape demonstrates how AI agents can be weaponized through 0-click attacks that are invisible to both users and conventional security methods, Bhavsar said.

Operant AI, a runtime AI defense platform, has discovered Shadow Escape, a zero-click attack that exploits the Model Context Protocol (MCP) used by connected AI agents — affecting platforms like ChatGPT, Claude, Gemini, and others. This is not a flaw in any single AI model or provider, but a new attack method that can target any AI agent or application using MCP, the company said.

As companies increasingly connect LLMs to internal tools and data through MCP, Shadow Escape reveals a hidden threat operating inside trusted systems and beyond the reach of traditional security tools. Operant AI estimates that trillions of private records could be exposed through such zero-click data leaks.

"While MCP has become a foundational protocol enabling powerful AI integrations, our research reveals that standard MCP configurations create unprecedented attack surfaces that operate beyond the reach of traditional security controls," said Vrajesh Bhavsar, CEO and co-founder of Operant AI.

Shadow Escape demonstrates how AI agents can be weaponized through 0-click attacks that are invisible to both users and conventional security methods, Bhavsar said.

Unlike traditional prompt injection or data leaks, this attack doesn’t need user error, phishing, or malicious browser extensions. Instead, it leverages the trust already granted to AI agents and AI assistants through legitimate MCP connections.