# Cisco unveils frontier AI red teaming and agent observability tools > Source: > Published: 2026-11-09 20:05:32+00:00 # Cisco unveils frontier AI red teaming and agent observability tools Photo: Tima Miroshnichenko / Pexels The networking giant is adding attack simulation and agent monitoring to its AI Defense platform as enterprises hand more work to autonomous software [Cisco](https://cryptobriefing.com/markets/cisco/) has introduced AI red teaming and agent monitoring tools aimed at securing the AI agents that enterprises are starting to deploy. The idea is to attack your own AI before someone else does, then keep watching it once it is live. The timing matters. Companies are moving from chatbots that answer questions to agents that take actions, and an agent with access to tools and data is a much bigger target than a model that only writes text. ## What Cisco is rolling out The new capabilities sit inside Cisco AI Defense, the company’s security platform for AI models and applications. Cisco’s version is algorithmic, meaning automated systems probe models and agents for weaknesses rather than relying only on human testers. The platform scans across more than 200 risk categories. Those categories include prompt injection and data exfiltration. Prompt injection is when an attacker slips instructions into an AI’s input to hijack its behavior. Data exfiltration is the polite term for sensitive information quietly leaving the building. The red teaming runs on multi-agent systems, so several AI attackers work together and adjust their tactics as they go. Cisco ties the testing to its Integrated AI Security & Safety Framework. That framework aligns with standards from NIST and MITRE. The monitoring side addresses a different problem. Testing an agent before launch is useful, but agents behave differently in the wild, and teams need visibility into what those agents are actually doing once deployed. ### AI, tech, and the markets they move—in one daily briefing. Daily. Free. Join 34,000+ readers across crypto, finance, and policy. ## How the platform got here Cisco AI Defense first launched in January 2025. The February 10, 2026 update was described as the biggest expansion since the platform debuted. It introduced the algorithmic red teaming capabilities covering agents and models. On March 23, 2026, Cisco launched a free, self-service Explorer Edition. Organizations can use it to evaluate AI models and agents in as little as 20 minutes, with no upfront cost. The Explorer Edition supports major agentic frameworks, the toolkits developers use to build agents. It also accommodates systems that connect agents to outside services and data. In June 2026, Cisco added Agent Validation and Agent Harness Testing to the Explorer Edition. These features target vulnerabilities specific to agents, including tool routes and content channels. An agent that reads emails, documents or web pages can be manipulated by malicious text hidden in that content, which is known as indirect prompt injection. Cisco’s newer tests probe for exactly that kind of attack through various content channels. ## Why agents change the security math A standalone model that gives a bad answer is embarrassing. An autonomous agent that follows a bad instruction can send files, call tools or take actions inside business systems, which is a different category of problem. Cisco is positioning AI Defense around a full lifecycle approach. Its framework follows a discover-detect-defend sequence: find the AI assets in use, spot the threats against them, then block those threats. The platform also pulls in real-time threat intelligence from Cisco Talos, the company’s security research group. That gives the testing a feed of current attack techniques rather than a static checklist that ages badly. Cisco has highlighted this strategy at various industry events, pitching it as a way to address emerging risks from frontier AI while helping enterprises trust their own deployments. **Disclosure:** This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our [Editorial Policy](https://cryptobriefing.com/editorial-policy/).