How SaaS Tools Enable Testing of AI Models and Agents

Cisco's DJ Sampath on Securing Agentic AI With Red Teaming and Guardrails Anna Delaney (annamadeline) • March 27, 2026

Enterprises could move from a world with 150 million developers to one with 3 billion to 4 billion people writing code as artificial intelligence tools and agents enable nontechnical users to build software, said DJ Sampath, senior vice president of AI software and platform at Cisco.

See Also: AI Impersonation Is the New Arms Race—Is Your Workforce Ready?

As enterprises move from pilots to production, ensuring security becomes critical when deploying mission-critical applications, Sampath said.

Agentic AI introduces a different operating model: These systems act persistently, execute tasks independently and interact with complex ecosystems of models, tools and third-party services, often without human oversight, increasing the attack surface.

As attackers exploit the expanded attack surface through prompt injection, model-specific exploits and compromised model repositories, enterprises must repeatedly red team models ...