Complete AI Security Platform - Guardian, Recon, and Layer

COMPANY OVERVIEW

End-to-end AI security platform delivering model scanning, red teaming, and runtime protection through unified Guardian, Recon, and Layer products. Secures AI applications from model selection and testing through deployment and runtime monitoring. Acquired by Palo Alto Networks in 2025 to integrate into Prisma AIRS platform. Powered by 17,000+ security researchers from huntr community and partnership with Hugging Face. Scanned 4.8M+ model versions, submitted 2,520 CVE records, and supports 8,000+ MLSecOps community members. Fortune Cyber 60 and CB Insights AI 100 recognition.

CORE FOCUS

Shifts AI security from reactive detection to proactive Secure by Design methodology across model lifecycle. Enables teams to select, import, test, and deploy models safely without sacrificing innovation velocity. Integrates threat research from huntr's 17K researcher community into automated scanning and testing workflows. Operates at enterprise scale with flexible deployment options, modular architecture, and cloud-native integrations. Partners with Hugging Face, AWS, Databricks, Microsoft, and Elastic to embed security into AI development pipelines. Addresses emerging threats through continuous model scanning, automated red teaming, and runtime defense against model-specific exploits like prompt injection and model poisoning.

PRODUCTS & TOOLS

Guardian – Model security scanner for safe model selection and import workflows.

• Scans models for malicious code, backdoors, and vulnerabilities before deployment
• Integrates with Hugging Face Hub to validate 4.8M+ model versions automatically
• Detects supply chain threats in open-source model dependencies
• Provides risk scoring and remediation guidance for vulnerable models
• Supports air-gapped and cloud deployments for flexible enterprise requirements

Recon – Automated red teaming platform for rigorous AI application testing.

• Fastest, most effective way to rigorously red team AI apps against adversarial attacks
• Simulates prompt injection, jailbreak attempts, and model manipulation techniques
• Generates test cases based on OWASP LLM Top 10 and MITRE ATLAS frameworks
• Continuous testing workflows validate model robustness before production release
• Integrates into CI/CD pipelines for automated security gates

Layer – Runtime protection platform with deep visibility and threat control.

• Stops AI threats instantly at runtime with real-time monitoring and blocking
• Detects and prevents prompt injection, data exfiltration, and unauthorized model access
• Provides complete visibility into model inputs, outputs, and inference patterns
• Enforces security policies across model deployments without code changes
• Logs all AI interactions for compliance audit trails and forensic analysis

huntr Community Integration – Crowdsourced threat intelligence from 17K+ security researchers.

• Largest community of AI security researchers discovering model vulnerabilities
• Submitted 2,520 CVE records for AI/ML vulnerabilities to MITRE database
• First- and third-party threat research feeds directly into Guardian, Recon, and Layer products
• Maintains 500+ specialized threat scanners for model-specific exploits

TARGET MARKET

Enterprises deploying AI applications requiring end-to-end model security from selection through runtime. MLSecOps teams managing model governance across open-source and proprietary AI systems. Organizations adopting LLMs, generative AI, and agentic workflows needing supply chain validation. Regulated industries (financial services, healthcare, government) requiring compliance audit trails for AI deployments. Cloud-native companies integrating AI into production applications seeking runtime threat prevention. Security teams responsible for AI red teaming, vulnerability management, and secure-by-design implementations.

DIFFERENTIATORS

Only unified platform securing full AI application lifecycle—model selection, testing, and runtime—eliminating gap-riddled multi-vendor approaches. Backed by 17K researcher community providing unrivaled threat intelligence unavailable to competitors relying solely on internal research teams. Partnership with Hugging Face delivers exclusive access to scan 4.8M+ model versions, largest validation footprint in industry. Palo Alto Networks acquisition integrates Protect AI into Prisma AIRS, combining network security expertise with AI-native threat detection. Secure by Design philosophy shifts security left into development workflows rather than bolting protections onto production deployments. Proven at scale with 2,520 CVE submissions and partnerships with AWS, Databricks, Microsoft, and Elastic validating enterprise-grade maturity.