At RSA Convention 2026 in San Francisco, Cisco introduced a sweeping set of safety capabilities designed to satisfy the speedy rise of AI brokers in enterprise operations.
“AI brokers aren’t simply making current work quicker; they’re a brand new workforce of affiliates that dramatically develop what organizations can accomplish,”
stated Jeetu Patel, President and Chief Product Officer at Cisco.
The headline bulletins embrace the extension of Zero Belief Entry to AI brokers by means of Cisco Id Intelligence, a brand new self-service safety testing software known as AI Protection: Explorer Version, an open-source agent safety framework named DefenseClaw, and a significant growth of AI-powered risk response capabilities inside Splunk.
Collectively, these symbolize one of the crucial complete makes an attempt but by a significant vendor to construct a safety basis purpose-built for agentic AI.
Breaking Down the Bulletins
The primary and most foundational announcement is the extension of Zero Belief Entry to AI brokers. Till now, this framework largely utilized to human identities. Cisco is increasing it by means of updates to its Duo id platform and Safe Entry product. Organizations can register brokers, assign them to an accountable human proprietor, and limit entry to solely the instruments and knowledge required for particular duties. An MCP gateway (a routing layer for agent-to-tool communications) ensures all agent exercise is seen and auditable, eliminating the blind spots legacy instruments usually create.
The second main announcement is AI Protection: Explorer Version, a self-service software that enables builders and safety groups to stress-test AI brokers earlier than they attain manufacturing. Constructed on the identical validation engine utilized by Cisco’s World 2000 prospects, it runs multi-turn adversarial simulations that mimic the sustained, manipulative interactions of a foul actor. It assessments resilience towards immediate injection, jailbreaks, and unsafe outputs. The software generates exportable safety reviews for compliance overview and integrates immediately into developer pipelines by means of GitHub Actions, GitLab, Jenkins, and different platforms.
Alongside this, Cisco is releasing an Agent Runtime SDK that embeds safety controls immediately into agent code through the construct course of and an LLM Safety Leaderboard that scores AI fashions on resilience to assaults, serving to organizations select which fashions to belief.
The third key announcement is DefenseClaw, an open-source safe agent framework designed to take away friction between growth and safety groups. It bundles a set of scanning and stock instruments together with Abilities Scanner, MCP Scanner, AI BoM, and CodeGuard, guaranteeing that each agent ability is scanned, each MCP server verified, and each AI asset cataloged mechanically. Cisco plans to combine DefenseClaw with NVIDIA’s OpenShell sandbox setting, automating safety checks that beforehand required guide steps or separate software installations.
The fourth pillar is the growth of AI-powered capabilities inside Splunk, Cisco’s safety operations platform. The headline addition is the Agentic SOC, a set of specialised AI brokers together with a Detection Builder, Triage Agent, Malware Risk Reversing Agent, and Guided Response Agent. These shift safety operations from a guide, reactive mannequin to 1 wherein routine investigation and response duties run autonomously at machine velocity. Supporting these features are new options resembling Publicity Analytics for real-time asset threat scoring, Detection Studio for managing the complete detection engineering lifecycle, and Federated Seek for correlating knowledge throughout a number of environments in a single question.
Why This Announcement Issues Now
Cisco’s ambition is grounded in a stark actuality the corporate uncovered in its personal analysis. Eighty-five % of main enterprise prospects are already experimenting with AI brokers, however solely 5% have moved them into stay manufacturing.
That hole isn’t about ambition or price range, it’s about belief. Cisco’s message is obvious: clear up the safety drawback, and adoption will speed up.
“Tasks shelved for lack of sources at the moment are inside attain. The one restrict is creativeness, and safety groups are the important thing to unlocking this chance by making the agentic workforce protected sufficient to belief,” Patel defined.
Governance has change into a recurring theme in conversations about AI brokers. Analysis exhibits the share of UK organizations actively deploying AI brokers has almost tripled in twelve months, from 22% to 62%. But 86% of leaders acknowledge that AI brokers introduce safety and compliance challenges current frameworks weren’t designed to deal with, and 85% consider deployment is outpacing conventional oversight strategies.
That is the hole Cisco is aiming to shut. By offering what quantities to a management aircraft for the agentic workforce, a centralized strategy to onboard, govern, monitor, and if obligatory halt AI brokers, Cisco argues that safety ought to not be the barrier conserving agentic AI in pilot mode or hindering its adoption.
The Greater Image
Cisco’s announcement at RSA 2026 displays a broader shift in how enterprise safety distributors are positioning themselves. The main target is not merely on defending people or methods from cyber threats however on governing a brand new class of actor working contained in the enterprise.
As Patel famous, AI brokers symbolize a brand new sort of workforce that requires onboarding, oversight, and accountability, identical to people in a zero belief setting.
The introduced capabilities vary from these already out there, resembling Detection Studio and the Malware Risk Reversing Agent, to options rolling out between April and June 2026.
For B2B know-how leaders assessing their AI technique, the central query is evolving. It’s not merely “What can AI brokers do for us?” however “Do we have now the infrastructure to control what they do?” Cisco is betting that safety would be the key to unlocking the agentic alternative at scale and that organizations mastering governance first will transfer quickest.
