Enkrypt AI Named a Specialist in the First Gartner® Emerging Market Quadrant for AI Application Security — Startup Vendors, September 2026

Gartner has published its first Emerging Market Quadrant for AI Application Security — Startup Vendors, and Enkrypt AI by Anaconda, is named a Specialist.

The report’s debut reflects a shift we’ve seen firsthand in how security teams think about AI.

The team started Enkrypt AI in July 2022, a few months before ChatGPT launched. Back then, security teams we talked to were still asking whether AI needed its own controls at all. Today they’re asking which controls to put in place before an AI agent ships.

How we read the market definition

Gartner defines the AI application security market as “technologies that protect enterprise-developed AI applications and agents by combining security testing, exposure management, and runtime defense to detect, alert, or block threats.”

In our work with customers, that definition brings three disciplines into one category, because no single one is sufficient on its own. Guardrails built without testing are guesses. Until you’ve attacked an agent, you don’t know which tricks it falls for, so you end up blocking the wrong things and missing the right ones. Testing without runtime defense gives you a report and no protection. And neither gives a compliance team what it needs most, which is evidence that the rules on paper are enforced in production.

The enterprises we start working with have written AI policies, but no system checks an agent’s actions against them.  In our experience, their needs show how quickly AI application security is shifting toward controls that enforce written policies every time an agent acts.

What we’re seeing in the field

In two months, we scanned 268,000+ tools across 25,000 Model Context Protocol (MCP) servers and found 143,000+ vulnerabilities, affecting 73% of those servers. Teams are connecting agents to these servers far faster than security teams can review them.

The risk has moved from what a model says to what an agent does. A manipulated chatbot gives a bad answer. A manipulated agent changes a record, shares a file, or sends a message. 

For example, instructions hidden in a web page or a document can trigger those actions without the user ever seeing them. The most damaging failures play out over several steps: a retrieved document nudges the agent’s plan, the plan picks a tool, and the tool takes an action nobody intended.

The choice of model affects risk too. Our red teaming has found exploitable risks across frontier AI providers and open source models. Models from different providers respond very differently to the same attacks. That’s why we maintain the LLM Safety Leaderboard, so teams can compare large language models (LLMs) before they commit.

What red teaming finds in practice

In one engagement, we red teamed a customer-facing hospitality chatbot. The bot confirmed a reservation that didn’t exist, recommended a competitor’s property over its own, and quoted different prices for the same hotel room depending on demographic signals in the conversation.

None of these were jailbreaks. The bot did what users asked, in ways that created brand, legal, and revenue risk the team hadn’t scoped as security issues. Those are the failures adaptive red teaming is built to find. When an agent can use tools and access data or services, unexpected behavior can go beyond a bad answer to an action the user never intended.

See how Enkrypt AI tests agents before release and protects at runtime.

Where Enkrypt AI fits

Enkrypt AI helps teams connect what they learn before an agent ships to the controls that protect it in production. 

• Agent Red Teaming tests agents against more than 300 attack categories, helping teams identify the guardrails and policies they need.
• Agent Guardrails and Agent Policy Engine enforce those controls at runtime and automatically record decisions, giving security and compliance teams ongoing evidence of policy enforcement.
• MCP Scanner and MCP Gateway extend testing and protection to the tools agents use.

Together, these capabilities help teams find risks before an agent ships, protect it as it acts, and show security and compliance reviewers which policies were enforced in production.

Where to start

If you have an agent on this quarter’s release list, red team it before it ships. The results show which guardrails and policies it needs, and give your security and compliance reviewers concrete evidence to sign off on.

See Enkrypt AI in action. Request a demo focused on your agent and tool stack to see how red teaming, runtime guardrails, and policy enforcement work together.

Gartner clients can read the full Emerging Market Quadrant for AI Application Security — Startup Vendors on gartner.com.

FAQs

What is the Gartner Emerging Market Quadrant?

Gartner describes it as “a visual snapshot, analysis and set of insights that describe an emerging technology market, examine key trends and highlight key providers and their capabilities using a two-by-two matrix.” Vendors are placed on two axes, Potential to Execute and Potential for Market Disruption, across four quadrants: Market Shapers, Pace Setters, Pioneers, and Specialists.

What did Gartner name Enkrypt AI by Anaconda?

In its first Emerging Market Quadrant for AI Application Security — Startup Vendors, published September 16, 2026, Gartner named Enkrypt AI by Anaconda a Specialist. Gartner clients can read the full analysis in the report (ID G00856640).

Has Gartner recognized Enkrypt AI before?

Yes. Gartner named Enkrypt AI a Cool Vendor® in AI Security in September 2025.

Why do AI applications need their own security?

Traditional application security assumes code behaves the same way on every run. AI applications respond differently depending on prompts, retrieved content, and tool responses, so certain risks only appear at runtime. Agents raise the stakes because they act on those inputs: an instruction hidden in an uploaded PDF can get an agent to issue a refund nobody asked for.

That’s why Enkrypt AI red teams applications and agents against 300+ attack categories before they ship, then turns those findings into the guardrails and policies that run in production. Each runtime decision is logged, so security and compliance teams have evidence their controls work

How does Enkrypt AI relate to Anaconda?

Enkrypt AI is part of Anaconda. Enkrypt AI by Anaconda delivers AI Security & Guardrails in the Anaconda Platform, one of four platform pillars alongside AI workspaces, AI orchestration, and AI artifacts.

Where can I read the full report?

Gartner clients can access it on the Gartner site.

Gartner, Emerging Market Quadrant for AI Application Security — Startup Vendors, Meghan Hollis, Dionisio Zumerle, Dennis Xu, Marissa Schmidt, 16 September 2026, ID G00856640.

GARTNER and Cool Vendor are trademarks of Gartner, Inc. and/or its affiliates.
Gartner does not endorse any company, vendor, product or service depicted in its publications, and does not advise technology users to select only those vendors with the highest ratings or other designation. Gartner publications consist of the opinions of Gartner’s business and technology insights organization and should not be construed as statements of fact. Gartner disclaims all warranties, expressed or implied, with respect to this publication, including any warranties of merchantability or fitness for a particular purpose.