Firewall for AI is here to help you protect your artificial intelligence apps, chatbots, and LLMs from emerging threats. Detect code injection and other issues in prompt entries, and flag improper language or data in your app's responses. Offered in limited availability.

Firewall for AI lets you:

• enforce responsible AI use with real-time inspection and policy controls
• detect prompt injection, other coding attacks, sensitive data, and harmful language submitted to your app
• detect improper response, data leakage, and toxic language in output from your AI app
• protect your AI apps wherever they live

Threats to AI apps

As more organizations adopt Artificial Intelligence (AI), they face new security threats that traditional Web Application Firewalls (WAFs) and API security tools are not equipped to handle, like:

• Prompt Injection Attacks. Manipulating AI-generated responses to leak sensitive information or bypass safety measures.
• Toxic Output & Hallucinations. AI-generated content that may include harmful, misleading, or offensive language.
• Data Exfiltration & Model Theft. Threat actors attempting to extract proprietary knowledge from AI models.
• Compliance & Governance Challenges. Increased regulatory scrutiny around AI-generated outputs and sensitive data handling.
• AI-Specific Denial of Service (DoS). Attackers overloading AI models with high-volume or malicious queries.

How Firewall for AI works

Firewall for AI is purpose-built to counter emerging threats that specifically target your LLM AI applications. It applies security guardrails to both inputs and outputs, and helps you secure AI chatbots, copilots, and virtual assistants by:

• detecting and blocking dangerous model inputs like prompt injections, jailbreaks, and remote code execution.
• ensuring AI-generated content adheres to your brand standards
• maintaining compliance with leading standards through governance and visibility
• mitigating data exfiltration and intellectual property theft from AI models
• protecting against AI-driven misinformation and bad output, including toxic, harmful speech, hallucinations, and more.

Firewall for AI closes the security gap that exists between your app and its LLM model. Here's what happens when a user enters a prompt in your AI app:

Set protections

Firewall for AI setup differs depending upon what you want to protect. If your AI app lives on:

• ​Akamai​'s Edge platform, you create a Firewall for AI configuration, where you specify what detections you want to use. Then link it to a security configuration that covers the website where your app lives, specifying where to detect.
• any other platform, you create a Firewall for AI configuration then you'll use this solution's API to set protections.

Read how in this solution's online help (login required).