Cisco’s advancement in cloud security has taken a significant leap with the introduction of new technology aimed at assisting developers in detecting and addressing vulnerabilities within artificial intelligence (AI) applications and their underlying models. The recent unveiling of the Cisco AI Defense offering on January 15 marks a milestone in the company’s efforts to combat data leakage risks associated with popular services like ChatGPT, Anthropic, and Copilot. This new technology has been made available to early-access customers, with plans for a general release scheduled for March.
The integration of AI Defense with Cisco Secure Access, the SSE cloud security portfolio launched last year, has expanded the capabilities of the offering. This software-as-a-service solution includes an array of security features such as zero-trust network access, VPN-as-a-service, secure Web gateway, cloud access security broker, firewall-as-a-service, and digital experience monitoring. The platform also provides administrators with access to the AI Defense dashboard through the Cisco Cloud Control interface, offering a centralized view of all cloud security offerings.
At the core of AI Defense is a focus on addressing security concerns related to AI implementation in organizations while also ensuring the safety and integrity of the technology. Jeetu Patel, Chief Product Officer and Executive VP at Cisco, emphasized the importance of maintaining both innovation and security in the fast-paced digital landscape. Citing data from Cisco’s 2024 AI Readiness Survey, which highlights the lack of readiness among organizations to prevent unauthorized tampering of AI and understanding threats specific to machine learning, Patel underscored the significance of AI Defense in addressing these challenges.
The key feature of AI Defense is its ability to validate AI models at scale, particularly targeting enterprise AppSecOps organizations. This validation process allows developers to ensure the integrity of AI models before they are deployed into production. Continuous monitoring and validation of AI models are essential, given the dynamic nature of these models as new data is introduced, changing the behavior of applications and agents. To enhance its capabilities in automated model verification, Cisco acquired Robust Intelligence, a startup known for its innovative approaches to identifying vulnerabilities in AI models through algorithmic red teaming and automated pen testing techniques.
Industry analysts have praised Cisco’s initiative in addressing automated model verification, acknowledging the company’s position as a leader in providing security solutions for GenAI. The acquisition of Robust Intelligence was seen as a strategic move by Cisco, demonstrating its commitment to staying ahead in the rapidly evolving field of AI security. As security threats targeting vulnerable AI models continue to rise, the introduction of AI Defense is expected to play a crucial role in safeguarding organizations against potential exploits and attacks.
Incidents like LLMjacking, where vulnerabilities in large language models are exploited through various means like prompt injections, supply chain attacks, and data poisoning, have highlighted the urgent need for robust security measures in the AI landscape. By leveraging the capabilities of AI Defense, organizations can proactively defend against such threats and ensure the safety and reliability of their AI applications. Cisco’s innovative approach to AI security sets a new standard in the industry, offering a comprehensive solution to address the complex challenges posed by evolving technology landscapes.