HiddenLayer AISec is an enterprise-grade AI security platform that provides comprehensive protection for machine learning models throughout their lifecycle.

Recognized by Gartner as a Cool Vendor for AI Security, HiddenLayer delivers unified supply chain security, runtime defense, posture management, and automated red teaming capabilities in a single platform.

What is HiddenLayer?

HiddenLayer addresses the growing security challenges that organizations face when deploying AI and machine learning models.

As enterprises increasingly rely on ML models for critical business decisions, these models become attractive targets for adversaries seeking to manipulate outcomes, steal intellectual property, or inject malicious code.

The AISec Platform 2.0 provides visibility into your entire AI ecosystem, detecting threats ranging from adversarial attacks and model poisoning to supply chain compromises.

HiddenLayer continuously monitors models from public repositories like Hugging Face, alerting you to newly discovered vulnerabilities before they impact your systems.

Founded by security experts with backgrounds at Cylance and the US intelligence community, HiddenLayer brings decades of threat research experience to the AI security domain.

Key Features

ModelScanner

ModelScanner automatically detects malicious code injections, pickle deserialization attacks, and architectural backdoors in ML models.

It scans models in 35+ formats including PyTorch, TensorFlow, ONNX, Keras, and GGUF.

The scanner identifies vulnerabilities before models enter your production environment.

Runtime Defense

Real-time monitoring protects deployed models from adversarial attacks, prompt injection, and inference manipulation.

The runtime defense layer detects and blocks attacks without introducing latency or requiring model modifications.

AI Bill of Materials (AIBOM)

Generate comprehensive inventories of your AI assets including models, datasets, and dependencies.

AIBOM enables compliance tracking, risk assessment, and supply chain visibility across your ML infrastructure.

Automated Red Teaming

Proactively test your models against known attack techniques aligned with MITRE ATLAS.

Automated red teaming identifies vulnerabilities in model robustness, data handling, and system integration before attackers discover them.

How to Use HiddenLayer

Console-Based Scanning

Access the HiddenLayer console to scan models directly:

Navigate to the Model Scanner section
Upload your model file or connect to your model registry
Review scan results and remediation recommendations
Export compliance reports for auditing

API Integration

import hiddenlayer

# Initialize the client
client = hiddenlayer.Client(api_key="your-api-key")

# Scan a model file
result = client.scan_model(
    model_path="./models/classifier.pkl",
    format="pickle"
)

# Check for vulnerabilities
if result.has_vulnerabilities():
    for vuln in result.vulnerabilities:
        print(f"Found: {vuln.type} - {vuln.severity}")

CLI Scanning

# Install HiddenLayer CLI
pip install hiddenlayer-cli

# Authenticate
hiddenlayer auth login

# Scan a model
hiddenlayer scan ./models/transformer.pt --format pytorch

# Generate AIBOM
hiddenlayer aibom generate --output aibom.json

Integration

CI/CD Pipeline Integration

Add HiddenLayer scanning to your ML pipeline:

name: ML Security Pipeline
on: [push]

jobs:
  model-security-scan:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - name: Setup HiddenLayer CLI
        run: pip install hiddenlayer-cli
      - name: Scan Models
        env:
          HIDDENLAYER_API_KEY: ${{ secrets.HIDDENLAYER_API_KEY }}
        run: |
          hiddenlayer scan ./models/ --recursive --fail-on high
      - name: Generate AIBOM
        run: hiddenlayer aibom generate --output aibom.json
      - name: Upload AIBOM
        uses: actions/upload-artifact@v4
        with:
          name: ai-bill-of-materials
          path: aibom.json

MLOps Platform Integration

HiddenLayer integrates with major MLOps platforms including:

MLflow: Automatic scanning of registered models
SageMaker: Model registry integration
Databricks: Unity Catalog scanning
Hugging Face: Continuous monitoring of model hubs

Compliance

HiddenLayer aligns with major AI governance frameworks:

NIST AI Risk Management Framework (RMF)
MITRE ATLAS adversarial threat framework
ISO 42001 AI management systems
EU AI Act compliance requirements

When to Use HiddenLayer

HiddenLayer is designed for enterprises that rely on machine learning models for business-critical applications and need comprehensive security coverage.

It fits organizations that require compliance with AI governance frameworks, want visibility into their AI supply chain, or need to protect models from adversarial attacks.

Consider HiddenLayer if you download models from public repositories like Hugging Face, deploy customer-facing AI applications, or operate in regulated industries where AI governance is mandatory.

The platform particularly benefits organizations with mature ML operations seeking enterprise-grade security without disrupting existing workflows.