Two carriers that have already lost customer data on a massive scale are now handing more of their networks to AI agents. That should make you pay attention. The pitch is faster repairs and fewer outages. The risk is a new set of doors into systems that hold your account details, location history and call records.
According to executive interviews with Reuters, as reported by Android Headlines, leaders at both companies confirmed they are boosting cybersecurity budgets to counter threats tied to AI automation. T-Mobile CTO John Saw said his team is increasing threat monitoring and pressing vendors to patch security holes quickly. Neither carrier gave a dollar figure, so “spend millions” is more a headline than a verified number. We also don’t know how much of the money is new and how much is relabeled existing spending.
The context matters. T-Mobile and AT&T have both had breaches that touched tens of millions of customer accounts. Asking the public to trust them with autonomous software, after those failures, is a big ask. Promises of better monitoring are the standard response after every incident, and customers rarely get a way to check them.
The automation is already real. T-Mobile uses self-organizing network tools to adjust antennas when nearby towers go dark. During Winter Storm Fern, those systems reportedly restored service to ninety percent of affected customers within eight hours, before local power grids recovered. That is a genuine benefit, and it’s hard to argue against during a disaster. But it is also the carrier’s own figure, and giving software the power to reconfigure towers means that software becomes a high-value target.
AT&T CTO Jeremy Legg pointed to a problem no single carrier can fix: third-party tools and autonomous agents. He questioned whether open-source developer communities hold themselves to strict security standards when they build tools that touch core cellular infrastructure. That is a fair worry. A compromised dependency or a poorly vetted agent can do damage that no firewall at the carrier will catch.
AT&T says it is responding with these measures:
- Heavy data encryption
- Multi-factor authentication
- Strict identity checks for external software agents
Those are baseline controls, not exotic ones. The identity check is the interesting part. Carriers now have to confirm whether an incoming agent actually represents a real customer, which is a hard problem when the requester is software acting on someone’s behalf. AT&T already uses AI to read trouble tickets and push automated fixes for basic bugs. Human engineers still review sensitive data requests, so rogue algorithms don’t reach private account databases. That human check is the most reassuring detail in the story, and also the one most likely to erode under cost pressure.
Both carriers say fully autonomous networks are still years away. Complex blackouts, broken hardware and advanced cyberattacks need experienced engineers making calls under pressure. I’d take that at face value for now. But the direction is clear: more automation, more software with privileged access, and more places for things to go wrong.
What should users do? Not much, realistically. You can’t opt out of a carrier’s network operations. You can use strong account PINs, enable port-out protection, and ask your carrier what data its AI systems can see. Regulators should ask the same question, and demand answers in writing, not in interviews.