Carriers & OperatorsAt TArtificial IntelligenceCybersecurityCritical Infrastructure
AT&T's Stankey: AI leaves everyone more vulnerable for up to two years
AT&T CEO John Stankey says AI has left every company more vulnerable, with a two-year fix backlog and industry sharing — not regulation — as the remedy.
Why it matters
- AT&T CEO John Stankey says AI makes everyone 'more vulnerable' and remediation will stretch out two years or more.
- The Alliance for Critical Infrastructure launched earlier this year with 50+ member companies; AT&T, Verizon, Charter, Comcast, Lumen and T-Mobile participate.
- Stankey favors cross-industry information sharing over government regulation, building on the US AI policy framework announced late last year.
- The UK NCSC describes AI as a 'force multiplier' raising the scale, speed and efficiency of cyber threats.
- Stankey cited Anthropic's Project Glasswing controlled rollout as a model for pre-release vulnerability assessment.
The story
AT&T CEO John Stankey put a hard number on how long he thinks it will take businesses to turn artificial intelligence from a security liability into an asset: up to 24 months, possibly longer, during which every company exposed to the technology is more at risk than before.
"At this moment in time, I think everybody's more vulnerable," Stankey said this week at Semafor's Silicon Valley and the World conference. He conceded that the admission is hardly breaking news, but it may be the starkest assessment AT&T's chief executive has offered of what AI is currently doing to his own network defenses and everyone else's.
The comments land as enterprises work out how quickly AI can accelerate the discovery and exploitation of network vulnerabilities, particularly in the wake of Anthropic's Claude Mythos model. The UK's National Cyber Security Centre has already framed AI as a "force multiplier" for cyber threats, one that raises their "scale, speed and efficiency" simultaneously.
Why the CEO rejects regulation as the answer
Stankey positioned himself as a realist rather than a pessimist — "I'm not a doomer," he said — and argued the risk can be managed back down. But his prescription is cooperation among critical infrastructure owners, government and AI developers, not new government rules.
That approach would, in his words, put the "meat on the bones" of the US AI policy framework announced at the end of last year. "The meat's going to be put on the bones by businesses and industry working with each other, working with those that are generating the new technology … And it's going to require better sharing," he said.
The gap in his argument is worth noting: faster information sharing across vertical industries is easy to endorse in principle, and Stankey offered no specifics on what telecom operators would actually exchange, with whom, or on what legal footing. Competitors in communications are being asked to pool threat intelligence at a speed that market incentives rarely reward.
What is the Alliance for Critical Infrastructure?
The venue Stankey named for that cooperation is the Alliance for Critical Infrastructure, which launched earlier this year with AT&T among its founding members. The group counts more than 50 companies across six sectors: communications, energy, financial services, information technology, transportation and water systems.
Stankey sits on its board alongside Verizon CEO Dan Schulman. Charter Communications, Comcast, Lumen Technologies and T-Mobile are also members — a roster that puts the chief executives of most major US broadband and wireless operators around the same table on security for the first time through this forum.
A 50-company alliance launched within the year has yet to prove it can triage AI-driven threats at anything like the speed those threats evolve. Stankey himself acknowledged the adaptation problem, saying industries need time to catch up to fast-moving AI attacks.
Project Glasswing as the working model
Stankey pointed to Anthropic's Project Glasswing, a controlled rollout program, as a template for how enterprises and AI developers can work together to assess security risks and find vulnerabilities before new models ship widely. The related precedent is SK Telecom's use of a gated Claude Mythos deployment to shield its core infrastructure.
Even with that model in place, Stankey was blunt about the backlog it exposes. "Now there's a long list of things that have to be addressed and fixed, many of which are going to stretch out for two years or more," he said. "The question, how much risk do you want to take during that period, and how do you mitigate that?"
That framing carries an implicit admission: the fixes are known, numerous and slow — a two-year window in which critical infrastructure operators are, by his own description, running elevated risk while the remediation queue clears.
The trust argument
Stankey closed with the commercial logic behind his position. AI, he argued, cannot simply be technology for technology's sake, because trust is the foundation the business case rests on.
"If the public loses trust, if an economy is interrupted … then we've got technology that is built on a bad foundation," he said. "So those of us who ultimately have commercial interests, those of us that have employees that we care about, those of us that operate in communities that we want to see be vibrant and stable, we should be trying to lead some of this stuff. And that's where I think business has to move forward."
It is a notable shift in posture from a carrier chief: leading on AI safety as a business necessity rather than awaiting mandates. Whether the Alliance for Critical Infrastructure can convert a 50-member roster and controlled-rollout pilots into measurable reductions in exposure over the next two years is the question Stankey leaves open — and one the group's first year of operation will begin to answer.
Also reported
Original: events.semafor.com
More from Rebecca Stone
Show full bio
Correspondent covering technology and product news in Carrier business, 5G and 6G, fiber and broadband at Telecom Gazette.
521 articles