Earlier this 12 months throughout Cisco Dwell Amsterdam, we introduced the largest replace to Cisco AI Protection since its preliminary launch. Whereas the replace included a wealth of recent capabilities, the underlying theme was singular: safety for AI brokers.
Within the months since, we’ve seen our clients undertake brokers at an unimaginable pace for a wide range of use circumstances: private assistants, engineering copilots, buyer assist, worker onboarding, AI provide chain optimization, and far, far more. Whereas AI Protection might assist companies develop and deploy these brokers securely, it was abundantly clear that no two brokers have been alike—and that our safety method needed to mirror that uniqueness.
Right this moment, we’re sharing the latest replace to Cisco AI Protection—and we’re getting private with agent safety. From the structure of our resolution to our algorithmic testing and runtime protections, this subsequent iteration of AI Protection is deeply customizable and context conscious. Whether or not your brokers are streamlining inside operations, delivering customized care to sufferers, or offering important banking providers to shoppers, their behaviors and dangers are distinctive. Now, with AI Protection, your safety can be too.
Let’s take a deeper have a look at what’s new with AI Protection.
Simulate and defend towards adaptive AI dangers
From the day we launched AI Protection, we’ve provided algorithmic purple group testing and runtime guardrails for the broad spectrum of threats going through AI methods. The fact stays that brokers are prone to distinctive dangers throughout totally different industries, functions, and deployment eventualities.
Right this moment, AI Protection is bringing customized, context-aware safety tailor-made to each agent with adaptive purple teaming and guardrails.
Adaptive purple teaming permits a consumer to supply customized targets for vulnerability testing their brokers. AI Protection will interpret these targets, consider the goal system, ideate, plan, and execute a classy multi-stage assault. Outcomes are analyzed to find out the assault’s feasibility and potential affect.
The identical customizability carries over into adaptive guardrails, which a consumer can construct within the all-new Coverage Studio. Merely describe the risk you need to defend towards in pure language and add any organizational coverage paperwork that is perhaps related. The Coverage Studio agent will ask follow-up inquiries to refine your coverage and guarantee exact safety.
Let’s have a look at a fast instance: you’re a monetary establishment utilizing agentic AI to assist clients with cash and asset administration. For compliance causes, you’re frightened about your agent offering prescriptive funding recommendation like inventory suggestions.
First, you leverage adaptive purple teaming in AI Protection to see if this can be a life like concern. Certainly, in a easy two-turn interplay, our simulation efficiently elicits inventory buying and selling recommendation. You progress to Coverage Studio and say, “Create a coverage to stop prescriptive monetary recommendation like inventory buying and selling ideas.” In flip, it asks you to think about a number of associated eventualities: hypotheticals, market information evaluation, definitions of economic merchandise, basic monetary steering, and extra. By defining these boundaries, you create a exact, extremely efficient guardrail to stop your agent from dangerous prescriptive recommendation whereas sustaining its usefulness as a buyer assistant.
Safe agentic provide chains seamlessly in growth
The personalization of an AI agent is pushed largely by its the important thing elements that represent its harness—the mannequin, instruments, abilities, file system, reminiscence and different sources it has entry to. An agent with a broad arsenal of capabilities, entry to inside information, and sweeping permissions will be extremely helpful—and equally harmful. As AI provide chains turn into more and more advanced, companies must steadiness utility with safety.
The newest replace to AI Protection makes it simpler than ever to strike this steadiness with out impeding the pace of innovation. Brokers and their full dependency graphs are now mechanically found throughout buyer codebases, cloud agent platforms, and container photos. Each asset is cataloged in a central AI stock and scanned for systemic vulnerabilities. Builders can set off these scans instantly from their CI/CD pipelines by way of the AI Protection CLI or SDK. By validating the safety of each agentic constructing block together with fashions, MCP servers, instruments, and abilities, AI Protection helps guarantee your brokers are safe by design.
A single compromised element can undermine the safety of a complete agentic system. Think about a healthcare group deploying an agent to help with affected person triage. AI Protection discovers that the agent relies upon on an MCP server with entry to affected person data and a third-party talent for symptom evaluation. A scan reveals that the talent requests broader permissions than vital, creating a possible information publicity threat. Every of those belongings is centrally seen within the AI Protection stock, and the talent vulnerability is highlighted for safety overview. This info equips your group to grasp potential blast radius and remediate points earlier than the agent goes stay.
Make AI Protection part of any agent platform
In the case of deploying AI Protection, we acknowledge that each buyer’s necessities are distinctive. Cloud environments, safety instruments, compute infrastructure, information sovereignty—each platform is deeply private.
Now, AI Protection leads the business with a very platform-agnostic method, defending brokers throughout any platform, cloud, or mannequin supplier. However infrastructure is barely half the battle; builders at this time construct brokers on frameworks like Amazon Bedrock AgentCore, Google Agent Growth Equipment, and LangChain. Every of those platforms has its personal tool-calling conventions, orchestration patterns, and belief boundaries to navigate.
AI Protection integrates natively with these agent ecosystems and all three main cloud suppliers. AI Protection additionally gives assist for NVIDIA accelerated computing within the cloud or for on-premises deployments as an integral element of the Cisco Safe AI Manufacturing facility with NVIDIA. This contains AI Protection integration with NVIDIA NeMo guardrails and NVIDIA AI Enterprise software program, in addition to with the open supply NVIDIA OpenShell agent harness.
Collectively, we’re bringing constant and strong safety wherever your brokers are constructed and run.
A complete, deeply private method to agent safety
Brokers are one of many defining transformative applied sciences of our time. With unprecedented functionality and autonomy, brokers promise each super potential worth and threat—the safety measures we put in place will in the end dictate which approach this goes.
On the finish of the day, agent safety has many aspects: community and infrastructure safety, identification administration, provide chain validation, purple group testing, runtime guardrails, and extra. Disparate options aren’t adequate—brokers demand a complete, deeply built-in safety method that’s purpose-built for his or her new threat panorama.
Cisco is combining many years of management in networking and safety with deep AI experience to ship complete agent safety in a approach solely Cisco can. With this newest replace to AI Protection, organizations can safe brokers throughout their platforms, functions, and distinctive operational necessities.
Over the subsequent few weeks, we’ll be sharing deep dives into these capabilities proper right here on our Cisco AI weblog. In case you’re becoming a member of us at Cisco Dwell Las Vegas, come go to the AI Protection sales space to expertise actually private agent safety for your self.
Some merchandise or options described could also be in varied levels of growth and provided on a when-and-if accessible foundation.