Empower Developers to Secure AI Applications Through Code

We understand that every organization's AI security needs and infrastructure are different. Palo Alto Networks wants to enable AI security in a manner that best aligns with those needs. Some customers have told us they want to stop threats at the network layer with no app changes, while others want to detect and prevent threats in app code without changing their network, and ​​some want to do defense-in-depth with both options. Now, that choice is yours with AI Runtime Security.

As organizations continue to build new AI applications and infuse existing applications with AI functionality, the risks of AI threats increase. However, all this growth comes with added risks. Hugging Face alone hosts over 1 million models, datasets and apps, and it has over 19 million users every month. Open-source technologies and a new wave of threats means your developers need to prepare to defend your AI ecosystem.

Earlier this year, Palo Alto Networks enabled infrastructure security teams to deploy a network layer enforcement to help secure AI ecosystems by protecting AI applications, models and data. Today, Palo Alto Networks is set to redefine AI security by offering AI security as code to our Palo Alto Networks product portfolio. In a market first, we are pleased to announce that AI Runtime Security is now available with both a network and API-based enforcement points.

AI Runtime Security-as-Code

Through this new API functionality, Palo Alto Networks brings AI security-as-code to market to enable AI security in a fast and easy way. Developers get access to a RESTful API and can embed the custom-generated code template within their existing application code. This enables the analysis of the payload within the prompts and responses users create between any developed AI apps and any models that power them. Customers send their prompts and model responses to the API in their app code and then receive a verdict. This indicates if a threat was detected, alongside the recommended action that should be taken.

To fine-tune their AI protection, organizations can define a security profile within Strata Cloud Manager and then apply those policies universally. These profiles enable security teams to control how certain threats are handled (allow or block), such as with prompt injections, malicious URLs and sensitive data leaks.

Flexible Deployment with Granular Security

Due to its lightweight and flexible form factor, the AI Runtime Security API also provides significant advantages for nimble, AI-focused organizations:

• Simple Deployment in Minutes - Customers with Strata Cloud Manager can start creating API keys and code templates for AI security within minutes.
• Security for Any Public or Private Model - Allow developers the flexibility to build AI applications around the needs of their infrastructure. ​​The fully model-agnostic API can be used to secure any app to model interaction in any environment.
• Secure AI Agent Development: Allow developers to secure AI agents on SaaS AI platforms using API calls from application code

Additionally, the AI Runtime Security API enables developers to get granular control of their AI ecosystem:

• Custom Error Responses - Allow users to understand why a prompt and/or response was blocked through custom error responses.
• Data Poisoning Scans - Ensure that the data fed to the model maintains its integrity. Developers can use the API to scan-batched training data for threats and poisoning before fine-tuning their models.

All of these benefits and features are available to current users of Software NGFW credits. Those customers are encouraged to take advantage of a free, 90-day trial of API-based Runtime Security today.

Learn More about AI Runtime Security

To learn more about the full scope of capabilities of AI Runtime Security, visit our product page. If you're ready to discuss how we can help protect your AI application ecosystem, contact us to see a demo and get access to a free trial.

Public link

Please use the above public link if you want to share this noodl on another website.