AI agents as the new network boundary: A security discussion

“Identity used to be the perimeter. With agentic AI, reasoning becomes the new boundary we must secure.” — Nagendra Bykampadi, head of security standards for Rakuten Symphony and O-RAN Alliance Security Work Group co-chair

As AI-native architectures reshape how networks operate, fixed perimeters are disappearing. On this week’s Zero-Touch Live, Rakuten Symphony CMO Geoff Hollingworth spoke with Nagendra Bykampadi about how autonomous agents are continuously extending attack surfaces, forcing a rethink of zero trust for the 6G era.

📺 Watch the replay now below.

Securing a world of sprawling autonomous AI agents

Nagendra explained the telecom security journey from hardware boundaries to identity-based zero trust, and now to a third frontier where the reasoning layer inside AI agents becomes the point of vulnerability. He outlined implications for operators and telco stakeholders as they prep agentic AI deployments in real networks, revealing that:

• Reasoning becomes the perimeter. When agents plan, decide and act, trust depends on whether their decisions stay aligned with operator intent and policy.
• Multi-agent systems create a living attack surface. Drift, poisoning, privilege inflation and goal deviation can emerge as agents learn from each other and adapt their behavior.
• Guardrails replace the firewall. Policies, safety checks and reasoning filters must operate at machine speed because humans cannot supervise autonomous logic in real time.

“Only autonomous systems can detect and respond to AI-speed attacks. Human oversight alone will never keep up,” said Nagendra.

What comes next

Nagendra closed by emphasizing that telecom is a critical infrastructure sector and cannot afford to learn on the job as agentic AI scales. Secure-by-design agents, adversarial testing, policy-driven reasoning and AI-based defense all need to be established up front as operators head toward 6G.

‍