AI Security Landscape

Score: 8.8/10

A focused runtime security layer protecting against prompt injection, PII leakage, and hallucinations via API.

Score: 8.6/10

A set of LLM safeguards designed to detect violating content across multiple use cases. Model-based guardrail.

Developer-friendly CLI tool for testing, evaluating, and red teaming LLM applications.

Score: 8.5/10

A GenAI-first platform focused on protecting LLM interactions, offering a secured gateway, browser integrations, and specialized protection for AI agents via MCP.

Platform for evaluating, logging, and refining AI products with enterprise-grade security and scale.

Score: 8.4/10

Unified platform for MLSecOps, focusing on model scanning, supply chain security (AIBOM), and runtime protection (Guardian).

Score: 8.4/10

Centralized platform enabling safe use of data and AI with strong governance and privacy controls.

Score: 8.4/10

Generative AI Red-teaming & Assessment Kit. Scans LLMs for hallucinations, data leakage, and prompt injection.

Open-source observability and analytics for LLM applications, focusing on traces and evaluations.

Score: 8.3/10

Secures the entire lifecycle of Generative AI, protecting employees from risky AI use and developers from insecure model integrations.

Score: 8.3/10

Leverages Zscaler's Zero Trust Exchange to provide visibility into Shadow AI, enforce data loss prevention (DLP) policies, and control access.

Score: 8.3/10

Acquired by Cisco, Robust Intelligence offers an AI firewall and model assessment platform to secure AI apps from development to production.

Score: 8.3/10

Specializes in PII identification and redaction for text, audio, and images, often used as a pre-processing layer for LLMs.

Score: 8.3/10

Open-source testing framework dedicated to ML models and LLMs, covering bias, performance, and security flaws.

Automated evaluation and security testing platform for Large Language Models to catch hallucinations and safety issues.

Score: 8.2/10

Integrated AI security platform providing visibility across the AI lifecycle, from development to production, ensuring compliant and secure model usage.

Score: 8.2/10

Designed to protect AI agents and Model Context Protocol (MCP) workflows through automated discovery, red teaming, and guardrails.

Score: 8.2/10

Acquired by Snowflake, TruEra provides deep diagnostics, testing, and monitoring for ML and LLM applications to ensure quality and reliability.

Score: 8.2/10

Comprehensive tool to fortify LLM security, offering sanitization, detection, and prevention of attacks.

Machine learning observability platform to monitor, troubleshoot, and explain model performance.

Score: 8.1/10

A suite including GenAI Protect, Application Protection, and Risk Scanner providing visibility and control over enterprise AI usage across browsers and apps.

Score: 8.1/10

An AI security layer that manages access controls, data masking, and audit logs for enterprise data connecting to LLMs.

Score: 8.1/10

A unified platform for monitoring, explaining, and securing ML models and LLMs, featuring a dedicated 'Trust Service' for guardrails.

Score: 8.1/10

Security and orchestration platform allowing enterprises to safely use public and private LLMs with rigorous policy enforcement.

Score: 8.1/10

Force-multiplies AppSec teams to design and deliver secure software, now with Agentic AI focus.

Score: 8.1/10

Open-source text metrics toolkit for monitoring language models, detecting quality and security issues.

Score: 8.1/10

Multi-layered defense against prompt injection attacks using heuristics, vector DBs, and LLM analysis.

Policy-as-code platform that now includes specialized authorization for AI agents and tool calls.

Cloud-native DLP platform that detects and redacts sensitive data in GenAI prompts and SaaS applications.

Score: 8/10

Unified platform for discovering shadow AI, assessing model risks (AI-SPM), and enforcing runtime protection.

Score: 8/10

Now part of Tenable, Apex Security provides visibility and risk assessment for AI models, focusing on the 'AI Exposure Graph'.

Score: 8/10

Observability and security platform for AI, offering 'LangKit' for telemetry and an AI Control Center for enforcing policy guardrails.

Score: 8/10

Observability and guardrails platform that ensures AI reliability by detecting hallucinations and enforcing policies in real-time.

Score: 8/10

End-to-end platform for automated security testing, runtime protection, and governance controls (Probe & Guard).

Score: 8/10

Facilitates secure application development and runtime protection, extending CNAPP to AI workloads.

Score: 8/10

Scans models (h5, pickle, saved_model) to determine if they contain unsafe code or malware.

Platform for evaluating, monitoring, and debugging LLM systems throughout the lifecycle.

AI-native data protection platform that provides visibility and control over sensitive data in GenAI prompts and RAG contexts.

Score: 7.9/10

Offensive-security platform automating adversarial testing for LLMs and custom agents to identify vulnerabilities before deployment.

Score: 7.9/10

Spun out of KPMG, Cranium focuses on AI Security Posture Management (AI-SPM) and generating AI Bill of Materials (AI BOM) for compliance.

Score: 7.9/10

A comprehensive platform for MLSecOps, offering model scanning (SAIF) and runtime detection (MDR) for adversarial attacks.

Score: 7.9/10

Part of the Arthur platform, Shield acts as a firewall to detect and block toxic, hallucinatory, or PII-leaking content.

Score: 7.9/10

Detects prompt injections and other LLM attacks. Can be used as a library or proxy.

Comprehensive AI monitoring and observability platform for computer vision, NLP, and tabular models.

Score: 7.9/10

Specializes in securing low-code/no-code platforms and AI agents. It focuses on 'Application Lifecycle Management' for agents, preventing data leakage and broken access control in Copilots.

Score: 7.8/10

End-to-end AI security platform offering AI Firewall, Usage Control, Agentic AI Security, and Automated Red Teaming for LLMs and Computer Vision.

Score: 7.8/10

Enables enterprises to increase productivity via GenAI with a native platform for visibility and control.

Score: 7.8/10

Protects enterprises from novel threats like indirect prompt injection and data exfiltration.

Score: 7.8/10

Platform offering an open-source AI gateway and automated red teaming for protection.

Score: 7.8/10

Automated adversary emulation platform protecting commercial and custom GenAI models, powered by dark web intel.

Score: 7.8/10

An open-source guard agent for AI agent runtime security, spanning personal to enterprise use. Promotes the AI-RSMS standard.

Platform for enforcing governance, compliance, and security policies across enterprise AI usage.

Score: 7.8/10

Provides '3D Runtime Defense' for modern stacks, protecting AI models and APIs in real-time without requiring code instrumentation.

Score: 7.7/10

Focuses on rigorous red teaming, offering a platform to simulate attacks on AI models to uncover vulnerabilities.

Score: 7.7/10

Offers 'Citadel Lens' for automated red teaming and evaluation of LLM applications, focusing on reliability and fairness.

Score: 7.7/10

Unified AI security layer providing visibility and guardrails across the organization.

Score: 7.7/10

Delivers 'Ascend AI' for pentesting and 'Defend AI' for visibility and guardrails.

Score: 7.7/10

Decompiles and analyzes Python pickle files to detect malicious code injection in ML models.

Score: 7.7/10

An open-source platform specifically designed to manage and secure Model Context Protocol (MCP) servers, providing a control plane for agent-tool interactions.

Score: 7.7/10

Extends security architectures to detect, analyze, and control AI use (Shadow AI and Embedded Agents) to prevent data loss and threat insertion.

Score: 7.6/10

Focuses on the entire AI lifecycle, securing the data science supply chain, runtime pipelines, and autonomous agents.

Score: 7.6/10

Provides a control layer to govern, secure, and monitor the use of LLMs within the enterprise, ensuring data privacy and compliance.

Score: 7.6/10

Protects the behavior of AI/ML and GenAI models at build time (testing) and run time (firewall).

Score: 7.6/10

Delivers comprehensive AI agent security, discovering agents and enforcing runtime guardrails.

Score: 7.6/10

A platform that monitors agent behavior in real-time to catch blind spots and steer agents toward safer actions using 'contextual agentic security'.

Score: 7.5/10

Provides runtime guardrails for RAG, LLMs, and AI agents, enforcing safety and privacy policies.

Score: 7.5/10

Scans outbound response traffic in real time for undesirable content and confidential data at layer 4.

Score: 7.5/10

A toolkit for adding programmable guardrails to LLM-based conversational systems.

Modern PAM solution that provides just-in-time access management for AI agents and humans.

A Python library for validating structures and data from Large Language Models. Excellent for ensuring JSON output.

Deterministic identity and access stack for AI agents, enabling per-task permission boxes.