Cisco Foundation AI Launches Agentic Security Systems to Protect Autonomous AI Environments

Cisco Foundation AI Tackles Growing AI Security Challenges

As artificial intelligence systems become more autonomous and embedded across enterprise environments, securing these AI-driven operations has emerged as a critical challenge. Cisco Foundation AI is addressing this need by introducing agentic security systems that combine reasoning capabilities, adaptive retrieval mechanisms, and human oversight to support real-world security operations at scale1

1

. The announcement includes three major releases: the Foundation-sec-8B-Reasoning model, the Adaptive AI Search Framework, and the PEAK Threat Hunting Assistant, all designed to protect autonomous AI systems while maintaining transparency and control2

2

.

Security teams are grappling with the challenge of protecting AI-driven environments that increasingly span cloud platforms, internal systems, and external data sources. Traditional security approaches struggle to keep up with the speed and complexity of modern AI workflows, particularly as agentic systems begin to make decisions and take actions independently2

2

. Unlike traditional AI systems that operate through single-step inference, agentic security systems are designed to pursue objectives over time, reason across multiple steps, adapt to new information, and interact safely with enterprise tools and data1

1

.

Foundation-sec-8B-Reasoning Enables Multi-Step Cybersecurity Analysis

The Foundation-sec-8B-Reasoning model represents the first open-weight reasoning model designed specifically for cybersecurity workflows. Unlike general-purpose language models, this model is optimized for multi-step cybersecurity analysis including threat modeling, attack path analysis, configuration review, and incident investigation1

1

. The model produces explicit reasoning traces alongside its outputs, allowing analysts to understand how conclusions are reached and supporting validation, trust, and regulatory requirements in high-impact security environments2

2

.

This transparency is essential for AI security operations where explainability and accountability to human operators remain paramount. The model is trained to reflect the analytical processes used by security practitioners, enabling structured analysis across tasks that correlate signals across logs, configurations, code, and threat intelligence over time1

1

.

Adaptive AI Search Framework Transforms Threat Intelligence Analysis

The Adaptive AI Search Framework extends beyond static query-based searches by enabling reasoning-driven information retrieval. Security analysis often involves navigating large, fragmented, and evolving information spaces where the relevance of data becomes clear only after intermediate findings are examined1

1

. The framework allows AI models to iteratively refine their search strategies as new information emerges, much like a human security expert would approach an investigation2

2

.

By supporting reflection, backtracking, and strategic query revision, the framework enables compact models to explore complex information spaces with greater accuracy and efficiency. This capability improves threat intelligence analysis, accelerates incident response, and supports proactive vulnerability research across diverse data sources1

1

. The framework is designed to adapt investigation paths when dealing with incomplete, noisy, or fragmented data sources, addressing a common challenge in modern security operations2

2

.

PEAK Threat Hunting Assistant Maintains Human Oversight in AI

The PEAK Threat Hunting Assistant demonstrates how reasoning and adaptive retrieval capabilities combine in practice for real-world security operations. This open-source agentic AI assistant automates threat hunting preparation by using teams of cooperating AI agents to research threat actors and techniques, analyze internal security data, and generate customized, step-by-step threat hunt plans2

2

. PEAK applies structured reasoning and adaptive retrieval to one of the most time-intensive aspects of security operations: threat hunt preparation1

1

.

Human oversight in AI remains central to PEAK's design. Security analysts guide the process, validate findings, and incorporate organizational context at every stage. With its bring-your-own-model optionality and user-controlled data access architecture, PEAK provides flexibility while maintaining enterprise governance and data security1

1

. This approach allows security teams to retain control over decisions, models, and data access while benefiting from automated preparation workflows2

2

.

According to Yaron Singer, vice president of AI and security at Cisco Foundation AI, these releases reflect how the company is "delivering disproportionate impact by addressing foundational challenges at the intersection of AI and security." Singer emphasized that "Cisco's approach emphasizes open, security-native foundations, enterprise deployability and architectural rigor," ensuring that security, transparency, and control are built into agentic systems from the outset2

2

. As agentic AI systems become central to enterprise operations, these tools position organizations to adopt AI with confidence while ensuring security remains foundational to their AI strategy.