Reviewed November 17, 2026
HiddenLayer logo
AI model and agent security platform

HiddenLayer review: company overview, platform and competitors

HiddenLayer is an Austin company that sells a platform for finding, testing and protecting AI models and agents, from model file scanning to runtime controls for coding agents. It is one of the more complete independent options left in the category, with real government and regulated-industry traction, but pricing is enterprise-only and most performance claims still come from the vendor.

Quick factsNovember 17, 2026
Our score: 8.0 / 10
AI model and agent security platform
Headquarters
Austin, Texas
Founded
2022
Owner
Private (venture backed; $100M Series B led by Delta-v Capital, September 2026)
Fits best
Large enterprises and public sector teams that run their own models or agents and want one vendor for scanning, red teaming and runtime defense, including self-hosted or isolated deployments
Products
AI Security Platform: AI Discovery, AI Supply Chain Security, AI Attack Simulation, AI Runtime Security, Agent Harness Security; professional services
Pricing
Quote only. A public AWS Marketplace listing sells full platform access on a 12-month contract; private offers are the norm.

Overview

What HiddenLayer does

HiddenLayer was started in 2022 by former Cylance staff, led by CEO Chris Sestito, after they saw machine learning models at their previous employer attacked. The company's pitch has stayed fairly constant since: AI systems are software assets with their own attack surface, so they need inventory, pre-deployment checks and runtime monitoring the way endpoints do. It raised a $50 million Series A in September 2023, co-led by M12 (Microsoft's venture fund) and Moore Strategic Ventures, and a $100 million Series B in September 2026 led by Delta-v Capital.

The product today is a single platform with four modules. AI Discovery builds an inventory of models, agents and AI workflows, including ones nobody registered. AI Supply Chain Security scans model files before they are deployed and produces an AI bill of materials. AI Attack Simulation runs adversarial tests mapped to MITRE ATLAS. AI Runtime Security sits inline in production to catch prompt injection, data leakage and abuse. In August 2026 HiddenLayer added Agent Harness Security, which applies runtime controls to coding agents such as Claude Code, Cursor and GitHub Copilot.

For a company overview, the useful context is that HiddenLayer is still independent while many of its early peers were bought: Protect AI went to Palo Alto Networks, Lakera to Check Point, Robust Intelligence to Cisco and Pangea to CrowdStrike. That makes HiddenLayer a neutral option for buyers who do not want their AI security tied to one endpoint or network vendor. It also means it now competes against platforms that can bundle AI security into existing contracts.

Strengths

  • Covers the full AI life cycle in one product: inventory, model scanning, attack simulation and runtime defense
  • Independent vendor that works across clouds, model providers and SIEMs
  • Air-gapped edition and Defense Department marketplace status suit government and defense buyers
  • Early runtime controls for coding agents, an area most rivals only started covering in 2026

Watch out for

  • No published pricing, and the public marketplace offer signals large-enterprise budgets
  • Bundled AI security from Palo Alto Networks, Cisco, CrowdStrike and Check Point can undercut it inside existing contracts
  • Detection quality figures are vendor-reported; independent test results are hard to find, so plan your own proof of concept
Is HiddenLayer right for you? Get a shortlist

Features

HiddenLayer feature by feature

Model scanning and AI supply chain

HiddenLayer scans model artifacts from sources such as Hugging Face, JFrog Artifactory and Databricks Unity Catalog before they reach production, looking for embedded malware, unsafe serialization and tampering. Results feed an AI bill of materials, which helps when auditors ask what third-party models you run. The vendor says scanning does not need access to training data.

Attack simulation and red teaming

The AI Attack Simulation module runs automated adversarial tests against models and agents, aligned to MITRE ATLAS, including system prompt evaluation and red team evaluation. It is useful for repeat testing before each release. Teams that need a human-led engagement can add HiddenLayer's professional services.

Runtime security for apps and agents

Runtime protection inspects prompts, responses and agent actions in production and can detect, redact or block. It covers prompt injection hidden in documents or tool responses as well as direct jailbreak attempts. Alerts can go to Splunk, Microsoft Sentinel or webhooks.

Agent Harness Security for coding agents

Launched in August 2026, this hooks into the native hook interfaces of coding agents such as Claude Code, Cursor and GitHub Copilot to flag prompt injection, sensitive data exposure and unsafe commands. HiddenLayer reports whether each action was enforced, only monitored or limited, because the controls available differ by agent.

Discovery and shadow AI inventory

AI Discovery maps models, endpoints, agents and AI workflows across the environment and flags ones that have not been reviewed. On AWS it covers Amazon Bedrock models and agents, SageMaker endpoints and agents built on Bedrock AgentCore.

Deployment choices

The platform is sold as SaaS, and HiddenLayer also offers an air-gapped edition through Carahsoft for classified or tightly regulated networks. In June 2026 it reached Awardable status in the Defense Department CDAO's Tradewinds Solutions Marketplace.

Pricing

HiddenLayer pricing

HiddenLayer does not publish a price list. Its AWS Marketplace listing is contract-based: you buy units of platform access for a set term, and the public offer shows a 12-month full-platform contract at $5,000,000, with private offers available on request. Treat that figure as a list ceiling, since most enterprise deals go through private offers or resellers and the listing does not define what one unit covers.

Expect pricing to scale with the number of models, agents and deployments protected and with the modules you take. Ask whether Agent Harness Security is licensed per developer seat or per agent, what the air-gapped edition adds, and whether professional services red team work is bundled or billed separately.

Alternatives

HiddenLayer alternatives

When another vendor may fit better.

PA
Palo Alto Networks Prisma AIRS

Pick it if you already run Palo Alto firewalls or Prisma Cloud, since it folds the former Protect AI model scanning and red teaming into that stack.

CA
Cisco AI Defense

Pick it if Cisco networking and Secure Access are your control points and you want AI validation and runtime guardrails from the same vendor.

CP
Check Point Lakera

Pick it if your main concern is prompt injection and data leakage in LLM apps and you want a lighter runtime guard rather than a full model security platform.

Noma Security logo
Noma Security

Pick it if you want AI security posture and agent governance tied closely to data science and development pipelines.

M
Mindgard

Pick it if continuous automated red teaming of models and AI apps is the main job and runtime blocking is a lower priority.

LS
Lasso Security

Pick it if you mostly need to control employee and application use of GenAI tools with fast setup.

Verdict

Should you buy HiddenLayer?

HiddenLayer is a strong fit for organizations that build or host their own models and agents, need self-hosted or isolated deployment, and want a vendor that is not tied to an endpoint or firewall platform. Smaller teams, or teams already paying Palo Alto Networks, Cisco or CrowdStrike, should compare bundled options first. Either way, run a proof of concept against your own models before signing a multi-year deal.

FAQ

HiddenLayer FAQ

Short answers about HiddenLayer.

Ask our research team
What is HiddenLayer?

HiddenLayer is a private AI security company based in Austin, Texas, founded in 2022. It sells an AI Security Platform that discovers AI assets, scans models for supply chain threats, runs automated attack simulation and protects AI apps and agents at runtime.

Is HiddenLayer an application security company?

Partly. It focuses on the AI layer of applications: models, prompts, agents and the files models are loaded from. It does not replace code scanning or traditional application security testing tools.

Who owns HiddenLayer?

HiddenLayer is privately held and venture backed. Its investors include Delta-v Capital, which led the $100 million Series B in September 2026, plus M12, Moore Strategic Ventures, Ten Eleven Ventures, Booz Allen Ventures, IBM Ventures and Capital One Ventures from earlier rounds.

How much does HiddenLayer cost?

Pricing is quote based. The public AWS Marketplace listing shows a 12-month full-platform contract at $5,000,000, but most customers buy through private offers sized to the models and agents they protect.

Who are HiddenLayer's competitors?

The main competitors are Palo Alto Networks Prisma AIRS, Cisco AI Defense, Check Point Lakera, Noma Security, Mindgard and Lasso Security. Large platforms now bundle AI security, so many deals come down to independent depth versus bundled convenience.

Can HiddenLayer run in an air-gapped environment?

Yes. HiddenLayer offers an air-gapped edition of its AI Security Platform, sold to public sector buyers through Carahsoft, for networks that cannot reach the internet.

Free and independent

Comparing HiddenLayer with other vendors?

Tell us what you need to protect and we will send three vendors that fit, including whether HiddenLayer is one of them.