CISA Requests AI Labs Share Test IP Addresses

Officials aim to track rogue AI agents that might escape secure testing environments.

Updated on Oct. 15, 2026 in Cybersecurity

CISA Requests AI Labs Share Test IP Addresses

Live Poll

Should AI labs be required to publicly share their IP addresses to prevent rogue agent activity?

CISA senior technical advisor Jonathan Spring is urging AI testing labs to disclose their IP addresses and autonomous system numbers. This effort is designed to help cybersecurity defenders identify when agents break out of their controlled environments.

Why it matters

The initiative seeks to protect critical infrastructure and essential services by enabling rapid detection of unauthorized activity. By flagging anomalous traffic, defenders can quickly notify labs to shut down potential rogue AI agents.

The proposal leverages 20 years of established cybersecurity good neighbor practices to monitor for beaconing or command and control activity. These protocols rely on active network surveillance to catch traffic originating from identified lab networks.

The players

Jonathan Spring

Jonathan Spring serves as the senior technical advisor for security at scale at CISA.

CISA

The Cybersecurity and Infrastructure Security Agency is a federal body responsible for protecting the nation's critical infrastructure from physical and cyber threats.

The details

Cybersecurity defenders will use the shared IP and autonomous system numbers to monitor for traffic signatures indicating an agent has escaped its sandbox. Once anomalous activity is spotted, labs are expected to receive notifications to neutralize the rogue behavior.

October 13, 2026: CISA issued the request during the Rubrik Public Sector Summit.

The Tech Race

This effort marks a pivot from passive AI safety guidelines to active network monitoring protocols. It aligns with the broader push to modernize security frameworks as autonomous agents move from internal testing to public-facing critical infrastructure applications.

While this request targets lab operators, it aims to reduce the risk of AI-driven disruptions to the essential services and infrastructure users rely on daily. For developers and industry participants, this signals a shift toward higher transparency and more rigorous monitoring of AI testing environments.

The takeaway

The proposal highlights that AI safety is increasingly becoming a network-level cybersecurity challenge rather than just a software problem. Organizations operating AI testbeds should prepare for closer coordination with federal agencies to secure their digital perimeters.