AT&T and T-Mobile Add AI Guardrails, but Humans Still Run the Network

AT&T and T-Mobile are using more artificial intelligence to manage their networks, but the same technology is forcing them to spend more on cybersecurity and keep people in control of critical decisions. Executives from both carriers told Reuters that fully autonomous telecom networks remain years away.

The risk is not simply that an AI model makes a bad recommendation. AI agents can act across several software systems with limited human input, which can expand the damage from a stolen credential, weak vendor control or poorly designed permission. A mistake could expose customer data or interrupt network operations rather than remain inside a test environment.

T-Mobile Chief Technology Officer John Saw said the carrier is increasing monitoring and threat response and working with vendors to close vulnerabilities faster. That caution follows major breaches in 2021 and 2023 that affected more than 100 million customer accounts in total. AT&T Chief Technology Officer Jeremy Legg pointed to risks outside a carrier's direct control, including third-party software and uneven security standards in open-source projects.

AT&T says its controls include encryption, multi-factor authentication, human review of data access and checks to verify that an outside AI agent is acting for a real customer. Those measures matter because an agent that can retrieve account information or change a service setting needs stronger identity and authorization checks than a chatbot that only answers questions.

Both companies already use AI in live network operations. AT&T applies it to review trouble tickets and sometimes carry out repairs. T-Mobile uses a self-organizing network to adjust cell-site antennas when another site fails. During Winter Storm Fern in January, T-Mobile said that system restored coverage to about 66% of affected customers within an hour and 90% within eight hours, often before electric power returned.

That result shows the business case for automation: a carrier can restore service faster and reduce the amount of routine work handled manually. It also explains why management is unwilling to remove people from the loop. Unusual failures, conflicting data and third-party dependencies are precisely the situations in which an automated action can spread a problem.

For investors, the development is primarily a cost and execution issue rather than a new revenue catalyst. AT&T spent $5.6 billion on capital investment in the third quarter, while it also works to extend fiber and wireless capacity. Security spending must compete with those priorities, yet underinvestment could create much larger costs through outages, customer losses, regulatory penalties and remediation.

The useful measure is not how often management uses the word AI. It is whether automation improves service reliability without increasing breach frequency or weakening accountability. AT&T and T-Mobile are moving toward more automated networks, but their expanding guardrails show that the safer path is controlled delegation, not unattended operation.