Anaconda Acquires Enkrypt AI After Finding 143,000 Vulnerabilities in MCP Servers

My take: Enkrypt AI scanned 25,000 MCP servers, the bridges that connect AI tools to the outside world, and found more than 143,000 vulnerabilities in 73% of them. This is not a theoretical study: it is an inventory of real gaps in the infrastructure many companies are already using to let their AI agents take automated actions.

MCP has become ubiquitous very quickly, and the speed of adoption always outpaces the speed of security hardening. When a technology grows this fast, controls arrive late. The problem is not the tool itself, but that it is being deployed without the governance it requires.

For any team building or deploying AI agents with access to external systems, this research is not a panic alarm: it is a reminder that AI governance includes securing every point of contact between AI and your data.

Do you know how many MCP servers your infrastructure relies on today, and who is auditing them?