Cisco Foundation Launches AI Security Tools for Autonomous Systems

The Cisco Foundation AI, a division of Cisco Systems Inc. dedicated to developing core artificial intelligence technologies, has announced an exciting range of new security tools aimed at bolstering agentic security. These tools are intended to assist enterprises in securing increasingly autonomous and interconnected AI systems integrated into their operations.

As organizations adopt AI across their business frameworks, the need for robust security measures becomes critical. These innovative tools empower AI systems to analyze intricate security issues, dynamically source evidence from various data sets, and execute security protocols—all while ensuring transparency and human supervision.

Security teams face the daunting challenge of safeguarding AI environments that now extend over cloud platforms, internal systems, and external data sources. Cisco Foundation AI asserts that conventional security methods are falling short in keeping pace with the rapid evolution of AI workflows, especially as agentic AI systems start to make independent decisions and take actions.

Among the new releases is the Foundation-sec-8B-Reasoning model, an open-weight reasoning model specifically tailored for cybersecurity applications. Unlike standard language models, this model is optimized for comprehensive security assessments, including threat modeling, attack path analysis, and incident investigation. It generates clear reasoning traces that help analysts comprehend how conclusions are derived, thus supporting validation, trust, and compliance in high-stakes security scenarios.

Additionally, Cisco unveiled the Adaptive AI Search Framework, a sophisticated information retrieval system that moves away from static search queries. This framework enables AI models to iteratively refine their search strategies in response to new findings, paralleling the adaptive strategies of human security professionals. It aims to enhance threat intelligence analysis and incident response by equipping AI systems to modify their investigative paths when dealing with incomplete or fragmented information.

The third announcement is the PEAK Threat Hunting Assistant, which is an open-source agentic AI assistant designed to streamline threat hunting processes. Utilizing teams of collaborative AI agents, it conducts research on threat actors and methodologies, analyzes internal security data, and creates tailored, step-by-step threat hunting plans. Cisco Foundation AI emphasizes that PEAK maintains human involvement, ensuring that security teams retain authority over decisions, models, and data access.

“Together, the reasoning model, AI search framework, and PEAK represent how Cisco Foundation AI is making a significant impact by tackling foundational issues at the intersection of AI and security,” stated Yaron Singer, Vice President of AI and Security at Cisco Foundation AI, in a blog post. “Our strategy focuses on open, security-native foundations, enterprise scalability, and architectural integrity.”

Image: Cisco