Cisco argues that enterprises should expect continued growth of AI agents and ever more capable AI models, and they should adopt integrated security and observability platforms to manage this evolving landscape.

Anthropic's Mythos frontier model can autonomously identify and exploit vulnerabilities, accelerating zero-day discovery and creating a dual-use risk for enterprises.

Key enterprise implications include compressed attack timelines, agent-driven surfaces that resist legacy security tools, and governance infrastructure for agents must be a first-class security consideration.

Cisco Cloud Control was launched to unify portfolio management, enabling natural-language queries and cross-domain threat correlation for security and observability.

Non-human identity is a core problem, prompting Cisco's acquisition of Astrix Security to govern machine identities and integrate agent governance with zero-trust platforms.

Cisco’s three-move response includes expanding AI Defense for agents with behavioral guardrails, revising zero-trust to emphasize action control for agents, and building an agentic SOC to triage alerts and predict breaches with AI.

In a keynote at Cisco Live, AI is framed as accelerating both defense and attacker capabilities, with a warning that the threat landscape is changing rapidly and the window to stay ahead is narrowing.

Observability is central to governance; Cisco’s acquisition of Galileo enables full-stack agent observability, including monitoring of agent behavior, token usage, and outcomes.

The rise of AI agents expands attack surfaces and network traffic, as agents operate autonomously with tool access, introducing risks like prompt injections and data poisoning.