Cisco Live 2025: Enhancing AI Security for Enterprises

At the recent Cisco Live EMEA event held in Amsterdam, Cisco unveiled a comprehensive security strategy specifically designed to facilitate the safe and efficient deployment of artificial intelligence (AI) within organizations. This initiative, known as Cisco AI Defense, aims to address the significant challenges associated with managing AI and large language models (LLMs).

As businesses increasingly adopt generative AI applications, they face various threats, including unauthorized use of AI technologies, model manipulation through prompt injections, and attacks targeting runtime environments. Cisco AI Defense seeks to mitigate these risks, enabling secure development, deployment, and operation of AI applications.

The core of Cisco's approach revolves around several protective measures categorized into four main components. Firstly, access to AI applications can be regulated through policy enforcement, preventing unauthorized usage. Secondly, an enhanced visibility into utilized AI cloud applications provides transparency regarding potential risks and threats. Continuous analysis of AI models will help identify security vulnerabilities and unexpected behaviors, while real-time protection mechanisms are in place to thwart attacks such as prompt injection and denial-of-service incidents. Cisco's own Threat Intelligence Network, Cisco Talos, will serve as a vital data source for detecting these attacks.

In addition to these protective measures, organizations can leverage Cisco Secure Access to enforce security policies for end-users. A notable application of this feature is to safeguard against unauthorized data leaks into public generative AI services. By defining which applications are permitted and which are blocked, companies can effectively prevent employees from inadvertently inputting sensitive corporate data into external AI models. Furthermore, the initiative includes provisions to protect proprietary and utilized AI models from potential manipulations through ongoing validation and red-teaming practices.

Alongside Cisco AI Defense, the company introduced the Hybrid Mesh Firewall, which merges traditional firewall functionalities with modern networking frameworks, offering granular protection for cloud and data center environments. The AI Defense framework is designed to be seamlessly integrated into this firewall.

Cisco AI Defense was initially announced in January 2025 and is set to be widely available starting March 2025. Concurrently, new Cisco N9300 Smart Switches featuring integrated data processing units (DPUs) are slated for release, enhancing security levels for AI workloads by efficiently distributing networking, storage, and security tasks at the hardware level. The Cloud On-Ramp Switch is expected to be available by April 2025, while the top-of-rack model is anticipated for summer 2025.