Cisco at RSAC 2026: Building Trust into Agentic AI
“We’re living in one of the most exciting and simultaneously the most disorienting times in human history,” stated Jeetu Patel, Cisco’s President and Chief Product Officer, during his keynote at RSAC 2026 in San Francisco this week.
Patel’s remarks underscore the unprecedented pace of change driven by agentic AI—autonomous software agents capable of independently executing complex tasks. While these technologies promise immense productivity gains, they also introduce new risks and uncertainties for organizations unprepared to deploy them responsibly. According to Patel, the key to navigating this landscape lies in trust—both in the AI agents themselves and in the frameworks that govern their operation.
The ability to delegate a task in a trusted form, versus just delegating a task without trust, is going to be the difference between being a market leader and being bankrupt,” Patel emphasized.
Cisco has committed to leading the charge in ensuring secure, trusted agentic AI deployments, particularly as enterprises prepare to manage thousands of independent agents operating across sensitive systems and data environments.
At RSAC 2026, Cisco unveiled several critical solutions and innovations designed to establish trust, security, and transparency in agentic AI. These include:
- Extended Zero Trust Access for Agents: Expanding Cisco’s Zero Trust security framework to provide granular identity management and access controls for AI agents.
- DefenseClaw: An open-source framework that safeguards agentic deployments, ensuring agents execute their tasks securely and as intended.
- AI Defense: Explorer Edition: A democratized toolkit that allows developers to assess AI safety and security comprehensively.
- Splunk Innovations: Including Exposure Analytics, Detection Studio, Federated Search, and expansions in the agentic Security Operations Center (SOC).
Patel stressed that these measures are essential to reimagining security for the agentic workforce, centered around three guiding strategies: protecting agents from external threats, protecting the broader ecosystem from rogue agents, and responding to incidents at machine speed and scale.
DefenseClaw: Protecting Agents from the World
One of Cisco’s flagship solutions, DefenseClaw, focuses on protecting agents from being compromised while operating in complex digital environments. With tools like OpenClaw accelerating agent deployment, the security imperative has grown.
Patel elaborated on the open-source approach:
We have a multitude of tools in the AI Defense Explorer Edition. The goal is to collaborate as a community, share knowledge, and make these tools freely available—from Skill Scanners to AI Bill of Materials to MCP Scanners, all hosted on GitHub.
DefenseClaw integrates these capabilities into a unified framework for OpenClaw deployments. It automatically activates services to scan agent skills, identify vulnerabilities, and monitor MCP servers, ensuring tasks are delegated in a trusted manner.
In addition, DefenseClaw is compatible with OpenShell, NVIDIA’s container for OpenClaw deployments, providing seamless integration for containerized AI workloads.
DJ Sampath, Cisco’s SVP and General Manager for AI Software and Platform, highlighted the lifecycle coverage of these tools:
AI Defense Explorer Edition lets you experiment and learn, while the enterprise edition connects to Cisco Secure Access, provides guardrails, and facilitates MCP scanning. This ensures the full AI lifecycle—from pilot to production—is secured against emerging threats.
Protecting the World from Rogue Agents: Extended Zero Trust
Securing the external environment from potentially rogue agents is another critical challenge. In the seminar From Chatbots to Change Agents: Securing Agentic AI, Matt Caulfield, VP of Product Management for Identity, and Kevin Kennedy, VP of Product and Solutions for Security, outlined strategies for identity and access governance in increasingly autonomous AI ecosystems.
As organizations deploy thousands of independent agents, the security risk profile changes dramatically,” Caulfield warned.
The Cisco approach emphasizes three critical steps:
- Know Your Agents: Organizations must first discover and inventory all agents operating within their environment. Without agent discovery, risks remain invisible.
- Control Access: Implement consistent enforcement points, such as AI gateways, to manage agent permissions and lifecycle policies effectively.
- Governance and Adaptation: Continuous monitoring and governance are essential, as authorized actions may still pose risks in dynamic contexts.
Kennedy added:
The key is to know your agents, authorize every action, and adapt to risk. Policy alone is insufficient—continuous oversight and real-time adjustments are critical.
From the customer perspective, Jeremy Nelson, CISO for North America at Insight, praised Cisco’s Zero Trust Access solution for AI agents:
Organizations are eager to embrace AI, but they need to do so without creating security gaps. Cisco’s Zero Trust Access for AI Agents provides visibility into agentic identities and restricts access to exactly what’s needed, securing data while scaling AI initiatives.
Securing the SOC at Machine Speed and Scale
Agentic AI also brings profound implications for Security Operations Centers (SOCs). Human analysts are limited by cognitive capacity, but agents can operate 24/7, at machine speed, and across vast datasets. This capability is both an opportunity and a risk, depending on governance.
John Morgan, SVP and GM of Splunk Security, and Fred Frey, Director of Software Engineering at Splunk, discussed how agentic AI transforms the SOC.
Threats are overwhelming analysts. Agentic AI empowers SOCs to operate at scale and prevent analyst burnout,” Morgan explained.
Frey elaborated on the concept of the Agentic SOC:
Agents need to understand business processes, data structures, and alert investigation workflows. They can retain and recall massive datasets, surfacing insights at the right moment. Trust is built as agents learn systems and processes.
Morgan concluded that, with appropriate trust and governance models, agentic systems become allies, enhancing efficiency and enabling proactive security.
An Agentic SOC is not about replacing humans—it’s about empowering them. Agents help predict, prevent, and respond to threats in a landscape where machine-speed attacks are the new normal.
A Comprehensive Approach to Agentic AI Security
Cisco’s RSAC 2026 presentations underscore a holistic approach to securing the emerging agentic workforce. From protecting agents and ensuring responsible delegation to safeguarding enterprises from rogue agents and enabling SOCs to operate at scale, the focus is on building trust and resilience across all touchpoints of AI deployment.
Patel summarized the vision succinctly:
The success of agentic AI hinges on trust. Without it, the potential is wasted, and risks multiply. With it, organizations can scale innovation safely and responsibly, unlocking unprecedented productivity and insight.
With these initiatives—DefenseClaw, AI Defense Explorer Edition, Zero Trust for AI agents, and Agentic SOC capabilities—Cisco is not only pioneering security for autonomous systems but also shaping a future where AI amplifies human potential safely.
Source link: https://newsroom.cisco.com
