Skip to content

DeepKeep

Tel Aviv-based AI security vendor offering a model-agnostic platform that scans, red-teams, and firewalls LLM and agentic AI deployments.

Visit Website ↗ + Add to Compare
60/100Incremental Innovator

Overview

DeepKeep builds a multi-layer AI security platform aimed at organizations deploying large language models, generative AI, and autonomous agents. The platform combines automated red-teaming, model scanning against a library of known vulnerabilities, an inline AI firewall, and usage-control policies intended to catch prompt injection, data leakage, and model manipulation before they reach production. It is deployable as SaaS, in a customer’s private cloud, or fully air-gapped, which has made it attractive to regulated and defense-adjacent buyers.

Founded in Tel Aviv in 2021, DeepKeep operates in the crowded AI/LLM security field alongside Protect AI, CalypsoAI, and Prompt Security, competing primarily on breadth of lifecycle coverage (pre-deployment scanning through runtime firewalling) rather than a single point capability. The company has raised a seed round led by Awz Ventures and drawn additional investment interest via OurCrowd, and has been featured in Cyber Defense Magazine spotlight coverage and named a winner in The Hacker News’ Cybersecurity Stars Awards for AI security.

As enterprises move from experimentation to production AI, DeepKeep’s bet is that continuous, lifecycle-wide monitoring will matter more than one-time model audits. The company is still early-stage relative to better-funded AI-security peers, and independent, named enterprise case studies are limited so far.

Innovation Matrix Assessment

Innovation Velocity 7/10

Rapidly expanded from model scanning into red-teaming and agentic AI security as the LLM threat landscape shifted.

Operational Value 6/10

Gives security teams a workable control layer for GenAI risk, though most enterprises are still early in AI governance maturity.

Market Momentum 4/10

Seed-stage funding and limited named enterprise deployments; momentum is real but modest next to larger AI-security rivals.

Category Disruption 6/10

Consolidates several AI-security functions into one platform, but the category itself is crowded with well-funded competitors.

Real-World Efficacy 5/10

Red-teaming and scanning claims are plausible but rely mostly on vendor-reported results rather than independent third-party testing.

Enduring Relevance 8/10

AI/LLM security will only grow in importance as agentic AI adoption accelerates across the enterprise.

Why CISOs Should Care

Gives CISOs a single control point for discovering, testing, and governing LLM and agent deployments before they become shadow-AI risk.

What Makes It Different

Full-lifecycle coverage — pre-deployment model scanning, automated red-teaming, and runtime AI firewalling — in one platform, deployable air-gapped.

The Matrix Verdict

60/100 — INCREMENTAL INNOVATOR

A credible, technically broad AI-security platform in an increasingly crowded field; worth watching as enterprise AI governance matures.

Editorial Note: Claims vs. Verified Findings

Efficacy and adoption figures are largely vendor-published; independent analyst or customer validation is limited.

Sources