# AI Security

MCP Defender is an AI firewall for monitoring and protecting MCP communications. It intercepts tool calls and responses, and verifies them against security signatures. MCP Defender provides advanced LLM-driven malicious activity detection and allows users to manage signatures used during scans.

AI Infra Guard is an AI infrastructure security assessment tool developed by Tencent. It focuses on discovering and detecting potential security risks in AI systems, supporting 28 AI framework fingerprint recognitions and covering more than 200 security vulnerability databases. The tool is lightweight, easy to use, requires no complex configuration, and features flexible matching syntax and cross-platform support. It provides an efficient assessment method for the security of AI infrastructure, helping enterprises and developers protect their AI systems from security threats.

Permit.io AI Access Control is a solution focused on AI application security and permission management. Through fine-grained access control, it ensures that every stage of AI interaction (from input to output) complies with security policies, preventing unauthorized input, data breaches, and harmful outputs. The product uses framework-agnostic technology, supports multiple AI frameworks, and offers flexible integration methods, allowing for gradual expansion into any AI technology stack. Permit.io aims to provide security for enterprise AI-driven applications, helping businesses enjoy the benefits of AI while meeting compliance requirements.

Verifiable Compute is an AI framework launched in collaboration with EQTY Lab, Intel, and NVIDIA, designed for governance and auditing of AI workflows. It represents a significant advancement in providing hardware-based solutions that ensure AI is interpretable, auditable, and secure during operation, instilling renewed confidence in accelerating AI adoption and development for consumers and enterprises. Verifiable Compute introduces a patented hardware-based encryption AI notarization and certification system to isolate sensitive AI operations and notarize every data object and piece of code computed during AI training and inference with immutable records. It also offers real-time compliance checks and enforcement of AI business policies, as well as adherence to new sovereign AI regulations, such as the EU AI Act. The new trust layer of Verifiable Compute is directly embedded in the silicon of next-generation hardware from NVIDIA and Intel, establishing new standards for AI safety and innovation.

Gecko Security is an AI-driven security engineer designed to discover and fix vulnerabilities in codebases. It leverages automated hacking instincts to create next-generation security tools, making it possible to find 0-day vulnerabilities. By verifying findings in a secure sandbox, Gecko Security reduces false positives, tests code like a hacker to uncover logical flaws that other tools might miss, and integrates into development environments to capture emerging vulnerabilities in real time. The product is developed by a team with security expertise, committed to enhancing code security without slowing down development speed. Pricing includes a free Open Beta version and customized plans for enterprise clients.

Protect AI is a comprehensive platform focused on AI security, offering end-to-end AI risk management and protective measures. It integrates AI Security Posture Management (AI-SPM) to help enterprises ensure the security of AI applications while safeguarding data and defending against unique AI-related threats. This platform provides comprehensive visibility, remediation, and governance for AI systems, allowing businesses to maintain confidence in AI exploration and innovation.

JailbreakZoo is a repository focused on breaking large models, including large language models and vision-language models. The project aims to explore vulnerabilities, exploitation methods, and defense mechanisms of these advanced AI models to promote a deeper understanding and awareness of security in large-scale AI systems.

DETECT-2B is Resemble AI's latest deepfake detection solution. It can detect over 30 languages with an accuracy of over 94% within 200 milliseconds. Our efficient and multilingual technology effectively addresses audio fraud based on AI.

ZeroTrusted.ai is a pioneering company specializing in generative AI security. Their LLM Firewall product is designed to protect you from the risk of data exposure and misuse by unethical language model providers or malicious actors due to potentially sensitive information in language model training datasets. The product offers anonymization features to protect prompt privacy, ensures data security and privacy through ztPolicyServer and ztDataPrivacy, optimizes prompts and verifies results to improve accuracy and prevent model hallucination, and supports integrations with various tools like LangChain and Zapier. The product comes in various pricing plans, including free, standard, business, and enterprise editions, each with different functionalities and service levels. ZeroTrusted.ai is committed to simplifying security compliance by utilizing cloud-agnostic Zero Trust solutions, dynamic adaptive encryption, and other technologies to maximize application and data protection.

PyRIT, a Python risk identification tool developed by Azure, is designed to assist security professionals and machine learning engineers in proactively detecting risks within their Generation AI systems. This tool automates AI red team tasks, allowing operators to focus on more complex and time-consuming tasks while also identifying safety and privacy hazards.

Lakera Guard offers enterprise-level security for machine learning models with just one line of code, preventing security risks such as prompt injection, data leaks, and generation of harmful content, allowing organizations to easily build secure and reliable AI applications.

NoPromptInjections is a product designed to protect AI applications from prompt injection threats. It offers a simple API interface for quick and easy use. Pricing is transparent, based on data size. Suitable for a variety of AI application scenarios.