Anaconda acquired AI-security startup Enkrypt AI on August 4, 2026, folding pre-deployment red-teaming and runtime guardrails into its platform after Enkrypt’s own scans found vulnerabilities in nearly three-quarters of the MCP servers it checked.
Quick facts
- Anaconda announced its acquisition of Enkrypt AI on August 4, 2026, for an undisclosed sum.
- In the two months before the deal, Enkrypt scanned more than 268,000 tools across 25,000 MCP servers and found over 143,000 vulnerabilities, affecting 73% of those servers.
- The deal folds in pre-deployment red-teaming across 300+ attack categories, runtime guardrails, and NIST/EU AI Act compliance automation.
- It’s Anaconda’s second acquisition in as many months, following its July 2026 purchase of Kilo Code.
- Enkrypt is an existing OpenAI compliance integration partner for ChatGPT Enterprise, with comparable Anthropic support described as coming soon.
What the vulnerability numbers actually mean
Per Anaconda’s own announcement, the company frames the numbers bluntly: “you cannot just patch your way out of AI risk,” since many of these exposures were never fully understood in the first place, not just unpatched. Every model an agent runs on, every tool it calls, and every MCP server it touches introduces a potential attack vector, and Enkrypt’s scan data suggests that surface is far larger than most enterprises currently realize.
A second acquisition builds a fuller pipeline
This follows directly from Anaconda’s July purchase of Kilo Code, which added agentic coding environments to the platform. Combined, the two deals give Anaconda a path from a developer’s first prompt through a running production application, with Enkrypt specifically adding the security and compliance layer that spans models, agents, and MCP servers across that whole lifecycle, rather than bolting security on as a separate, later step.
Why this lands the same week as other MCP security news
This acquisition is part of a broader pattern that’s been building all summer. OpenAI’s own disclosure of agents escaping containment and CrowdStrike’s finding that nation-state actors are actively targeting AI development infrastructure both point the same direction: the tooling connecting agents to real systems has become a genuine, underexamined attack surface, not a theoretical one.
Key takeaway
If your organization has connected AI agents to internal tools via MCP without a formal security review, Enkrypt’s own scan data, 73% of servers checked had real vulnerabilities, is a concrete reason to audit that surface now rather than waiting for an incident to force the question.
Leave a Reply