The adoption of generative AI (GenAI) in customer experience software is reshaping how organizations engage with their customers. From self-service chatbots to agent assist tools, personalization in marketing, and predictive analytics in sales, GenAI promises transformative capabilities. However, while GenAI opens doors to innovation, it also introduces unique vulnerabilities that must be addressed to ensure trust, security, and reliability. We recently published a comprehensive report on GenAI Trust and Safety that can obtained courtesy of Parloa, providers of the AI Agent Management Platform. 

As we integrate GenAI into customer-facing operations, several risks emerge. Large Language Models (LLMs), the backbone of many GenAI applications, are probabilistic tools. They generate responses based on patterns in data, which means they can inadvertently produce incorrect or misleading outputs, commonly referred to as “hallucinations.” Furthermore, these models are susceptible to adversarial attacks, such as jailbreaking attempts, where malicious users exploit the model to bypass safeguards and elicit inappropriate responses. To mitigate these risks, a robust defense strategy is essential. 

Enter Cisco AI Defense 

Cisco’s newly announced Cisco AI Defense product is designed to tackle these challenges head-on. This comprehensive tool offers organizations a way to monitor and secure their GenAI applications at the network layer. It also offers robust testing tools. Here’s how it works: 

• API-Level Integration: By obtaining an API key and connecting it to the network over which API calls to LLMs are routed, the AI Defense application intercepts all incoming and outgoing traffic. This enables it to enforce guardrail policies across the entire communication channel. 
• Real-Time Interception and Response: If a user attempts to jailbreak an LLM-powered application, AI Defense detects the attempt and ensures the LLM responds with predefined guardrail messages, such as “Sorry, I can’t help with that.” If an LLM is tricked into offering sensitive data, the misstep can be stopped before the data is leaked. 
• Algorithmic Redteaming: AI Defense also offers “algorithmic redteaming,” a feature that simulates a wide range of potential attacks or exploitation attempts in a controlled environment. The insights from these simulations allow organizations to harden their applications against vulnerabilities before they are exposed to real-world users. 

Implications for CCaaS Vendors 

In a pre-announcement demo offered to industry analysts, Senior Vice President of Cisco Security Product Management Raj Chopra stated that CCaaS vendors are ideal customers for Cisco AI Defense. CCaaS providers are increasingly embedding LLM-based chatbots and agent-assist tools into their platforms. By integrating AI Defense, CCaaS vendors can offer their customers added assurance that their GenAI solutions are equipped to detect and mitigate malicious inputs and problematic outputs at the network layer. 

This approach aligns with a growing need for GenAI applications to not only deliver value but also inspire trust. CCaaS vendors already recognize the importance of building guardrails directly into their platforms. For those seeking a partner to augment these capabilities, Cisco’s AI Defense provides an additional option. 

A Competitive Landscape 

It’s worth noting that Cisco is not alone in addressing the security challenges of GenAI. Several specialty vendors, covered in our recent GenAI Trust and Safety whitepaper, offer similar capabilities. These vendors have pioneered tools to address risks associated with LLMs, including malicious exploitation and response validation. Cisco’s entry into this space underscores the growing importance of these defensive measures as GenAI becomes a standard component of customer experience technology. 

Building a Safer Future for GenAI 

The benefits of GenAI in customer experience are undeniable, but the risks cannot be ignored. As we continue to leverage these powerful tools, it’s critical to adopt strategies that ensure their safe and effective use. Cisco AI Defense offers a new a robust solution to assist organizations in proactively addressing the vulnerabilities of GenAI while maximizing its potential. For CCaaS vendors and other enterprises integrating LLM-based solutions, the question is no longer whether to invest in GenAI but how to secure it. With tools like Cisco AI Defense, businesses can strike a balance between innovation and security, paving the way for a future where GenAI enhances customer experiences without compromising trust. 

If you’ll be attending Enterprise Connect in Orlando this March, I invite you to attend my presentation on this topic.

Categories: Articles