AI Security Security Tools

Oligo Security offers application layer insights as part of a CADR platform. They baseline application library activities at a function level, and can detect either malicious deviation, or the execution of known vulnerable functions. This extends into AI, allowing them to see, detect, and respond to AI applications.

Upwind has built CNAPP from the ground-up around runtime insights provided from a network focused endpoint agent. They provide unique features like API security, alongside vulnerability prioritization and scanning, and the more generic CNAPP feature sets.

The core of Sweet Security is powerful incident detection and response capabilities for cloud environments. They have since created a larger CNAPP platform expanding more into vulnerability management, posture, and identity capabilities

Raven has built a comprehensive runtime oriented ADR solution that can detect function executions from packages being exploited. This empowers them to detect application layer attacks, create prioritization based on what functions are being used, as well as virtual patching to prevent vulnerability exploitation.

Noma's AI Security solution focuses on a broad range of implementations and use cases across posture and runtime protection. They're especially strong for companies looking to secure the work of machine learning or AI engineering teams more broadly.

Pillar provides complete coverage as a developer first AI security solution. They integrate with your SCM, detect & diagram your AI agents, test them for vulnerabilities, and provide real-time protection and guardrails.

Lasso offers a combination of solutions to offer governance over AI solutions - from plugins to endpoint agents for monitoring and blocking.

Operant has focused their application defense solution on securing AI workloads, providing features like runtime detection and redaction for production applications. They excel at securing Kubernetes workloads, and are especially strong at network detections.

Snyk was the first to really do DevSecOps - fast, actionable, developer focused CI/CD security scanning. They also were early to consolidate scanners into a single place - IaC, Code, SCA, and Container. There are many platforms with small improvements over them - better reachability, workflows, etc. - but they remain a strong player in the space for having scaled so meaningfully across enterprises.

Akto has created an open source flavored approach to next generation DAST and API security with features like looking at log data for API discovery, sensitive data flows, and customized scanning. A uniquely helpful feature is the ability to easily edit and tweak tests from the UI.

Miggo is maximizing the value of your existing application performance monitoring, or offers an instrumentation of their own, that excels at building maps of distributed systems and real time attack detection and prevention. Miggo has no-code as well integration based implementations.

Acquired by Checkpoint. Lakera offers a simple way to protect LLM's by importing their SDK into your code. Their approach is simple and elegant, and their Gandalf tool allows you to better understand how LLM prompt injections work.

Acquired by Crowdstrike. Pangea provides an API and SDK to easily implement security features into your application - such as checking a user's password against a breach database, or checking a user's email against a spam database. They've expanded the capabilities of the SDK to focus on runtime AI protection.

Acquired by SentinelOne. Prompt Security offers comprehensive solutions for LLM security. They have both corporate IT visibility with their browser plugin, alongside application visibility with API, SDK, and reverse proxy options. You can also trace user sessions and detect/redact/block numerous types of data and attacks.

TrojAI deploys on premise and provides end to end LLM security - providing DAST type testing for LLMs and realtime protection via either a reverse proxy or SDK.

Mindgard has taken a cool approach to LLM security by building an in depth testing library for your existing models. Given the rapidly changing nature of the field, it's a great way to learn about existing attacks and how to protect against them.

Marqus focuses on runtime security for LLMs.