Radware Launches LLM Firewall to Secure Enterprise Generative AI Usage

1. 
   Home
   
2. 
   News
   
3. 
   RDWR
   
4. Radware Protects Generative AI Use within Enterprises with New LLM Firewall

11/18/2025 – 06:01 AM

Radware (NASDAQ: RDWR), a cybersecurity and application delivery solutions provider, today announced the launch of its LLM Firewall, an add-on to its Cloud Application Protection Services. This product represents Radware’s initial foray into agentic AI protection for enterprises. The LLM Firewall aims to address the security challenges arising from the growing integration of Large Language Models (LLMs) into applications, protecting LLM prompts and responses from attacks and misuse.

The LLM Firewall leverages real-time, AI-based protection at the prompt level to secure generative AI usage, effectively stopping threats before they can reach the LLM. Its model-agnostic design is intended for seamless integration across various platforms, allowing enterprises to secure AI use without disrupting workflows or innovation.

“Organizations are rightly concerned about adopting AI due to complex regulations, data safety and systems integrity,” said Constance Stack, chief growth officer, Radware. “Our new LLM Firewall is designed with the security enforced at the prompt in order to defend against prompt injection, jailbreaks, and resource abuse. Positioned as a Web Application Firewall (WAF) for LLMs, it mitigates natural language exploits, enhancing protection for LLM models and integrations in real-time.”

This solution is engineered to detect and block attempts to exfiltrate personally identifiable information before it interacts with the LLM, thus helping prevent the exposure of sensitive data. This capability can support compliance with global standards such as GDPR, HIPAA, and other data protection frameworks and enterprise policies. Radware has stated its LLM Firewall is designed to address the 2025 OWASP Top 10 Risks and Mitigations for LLMs and Gen AI Apps.

From a technical standpoint, the LLM Firewall likely employs a combination of techniques, including:

• Prompt analysis: Examining user inputs for malicious intent, such as prompt injection or jailbreaking attempts. This may involve natural language processing (NLP) techniques to identify suspicious patterns and keywords.
• Content filtering: Preventing the transmission of sensitive data by identifying and blocking prompts or responses that contain personally identifiable information (PII) or other confidential data.
• Rate limiting: Restricting the number of requests that can be submitted to the LLM within a given timeframe, preventing resource abuse and denial-of-service attacks.
• Behavioral analysis: Monitoring the interaction between users and the LLM to detect anomalous behavior that may indicate a security threat. This could involve identifying unusual patterns of prompt usage or unexpected responses from the LLM.

While the offering is positioned as a “WAF for LLMs,” the challenge lies in the nuances of natural language. Unlike traditional WAFs dealing with structured HTTP requests, LLM Firewalls must understand and interpret the semantics of human language and the potential for manipulation. The effectiveness will depend heavily on the sophistication of its AI-driven detection algorithms and its ability to adapt to evolving attack vectors.

Radware’s move highlights the growing need for specialized security solutions as generative AI becomes increasingly integrated into enterprise applications. As LLMs mature and become more powerful, the risks associated with their use will continue to grow, necessitating robust security measures to protect against both intentional attacks and unintentional data leaks.

For further details on Radware’s Cloud Application Firewall Services, visit [Radware Cloud Application Firewall Services].

Radware (NASDAQ: RDWR) is a global leader in application security and delivery solutions for multi-cloud environments. For more information, please visit the Radware website.