Telecom operators are automating more network and business processes, but a recent warning from the industry highlights an uncomfortable reality: the more authority AI receives, the more important operational security becomes.
On October 23, Reuters reported that AT&T and T-Mobile are strengthening safeguards as they expand their use of AI. The concerns include sensitive data exposure, vulnerabilities in third-party software and the possibility of AI-driven actions disrupting network operations. AT&T has introduced safeguards such as encryption, multifactor authentication, human oversight and verification of AI agents requesting customer account access.
The implications extend beyond AI itself.
Consider an automated workflow that investigates a billing anomaly, checks a subscriber’s service status and initiates a correction. Each step may involve a different system, permission and source of customer data. If the AI agent has excessive access, incomplete context or no reliable record of its actions, a process designed to save time can introduce a new operational risk.
This is particularly relevant as telecom companies connect CRM, billing, provisioning, network management and customer support through APIs. Integration makes automation possible, but connectivity between systems must not mean unrestricted access between them.
A more sustainable approach is to give automated systems narrowly defined permissions, validate actions before execution, maintain audit trails and require human approval for high-impact changes. The objective is not to slow automation down. It is to make automation dependable enough for production use.
For MVNOs and MVNAs, which often rely on multiple technology partners, these controls are just as important as speed of deployment. A modern BSS/OSS platform must support efficient operations while preserving clear boundaries around customer data, billing changes and service provisioning. This is an important consideration for API-driven platforms such as TelcoEdge as telecom automation expands.
The industry has spent years discussing how to make telecom systems more intelligent. The next challenge is making sure those systems can act safely, consistently and accountably.
Where should telecom operators draw the line between autonomous execution and human approval, particularly when AI can change billing, subscriber access or live network configurations?