Fortifying Your AI: A Developer’s Guide to Now Assist Guardian and Data Privacy

Now Assist Guardian and Data Privacy for Now Assist are two powerful features providing the necessary guardrails to ensure that your AI-powered solutions are not only innovative but also safe, secure, and compliant.

Now Assist Guardian: Your AI’s First Line of Defense

Now Assist Guardian is a built-in security and compliance layer that acts as a vigilant protector for all your Now Assist interactions. It is designed to mitigate the risks associated with AI-generated content by detecting and managing offensive language, security threats, and sensitive topics. It is model-agnostic, meaning it works with any large language model (LLM) you use on the Now Platform.

How it Works

At its core, Now Assist Guardian monitors both the prompts sent to LLMs and the responses they generate in real time. This analysis allows it to identify and flag potentially harmful content based on a set of configurable guardrails across various categories. For detected issues, you have the flexibility to either log the event for review or block the content entirely, preventing it from ever reaching the end-user.

A key capability is its ability to thwart prompt injection attacks, which are malicious attempts by users to manipulate the AI’s behavior. By recognizing these security exploits, Guardian can block the suspicious requests and log the attempt for administrative review.

Sensitive Topic Detection

Now Assist Guardian can identify conversations related to sensitive workplace issues, such as harassment or discrimination, and intelligently redirect users to the appropriate channels, like a live HR representative, rather than allowing the AI to handle them. This ensures that delicate matters are managed with the necessary human touch and in compliance with company policies.

Integration and Monitoring

Now Assist Guardian is seamlessly integrated with the Generative AI Controller and can be configured from the Now Assist admin console. It provides dashboards to monitor and analyze the data it collects, offering insights into detected threats and the performance of your AI applications. Now Assist Guardian is included with your Now Assist licensing, but the sensitive topic detection feature requires a Now Assist for HRSD license.

Data Privacy for Now Assist: Safeguarding Your Most Valuable Asset

Data Privacy for Now Assist is your primary tool for protecting sensitive information when leveraging generative AI. Its main purpose is to prevent the exposure of Personally Identifiable Information (PII) and Protected Health Information (PHI) in both the data sent to and received from LLMs.

How it Works

The feature uses real-time anonymization to mask sensitive data before it leaves your instance. Data is encrypted in transit using TLS 1.2 and is processed in-memory, meaning it is deleted immediately after the AI generates a response and is not stored or cached. ServiceNow provides out-of-the-box data privacy patterns, and you also have the flexibility to define your own custom patterns to control what gets masked.

The Role of the Sensitive Data Handler

For those familiar with the Sensitive Data Handler, it’s important to understand its relationship with Data Privacy for Now Assist. The Sensitive Data Handler is a tool that detects and masks sensitive data in real-time within your instance across conversational interfaces like Virtual Agent and Agent Chat.

Data Sharing and Control

ServiceNow is transparent about its data handling practices and provides customers with the ability to opt-out of sharing their data for AI model improvements via the Now Assist Admin Console. This gives you complete control over your data and how it’s used. For customers who do opt-in, ServiceNow uses industry-standard and in-house methods to cleanse PII and other sensitive details before the data is used for model improvement.

A Unified Front for Responsible AI

Now Assist Guardian and Data Privacy for Now Assist are two integral components of a comprehensive AI governance framework. While Guardian focuses on the content and behavior of your AI, Data Privacy for Now Assist is dedicated to protecting the data that fuels it. Together, they provide a multi-layered defense that enables you to innovate with confidence, knowing that your AI is both powerful and principled. By embracing these tools, you are not just building applications; you are building trust.