AI in Telecommunications
The New Attack Surface
My talk given at the Adastral Park,BT Group Advanced Connectivity Technologies (ACT)Now 2026 event covered how AI is rapidly becoming embedded in telecommunications networks.
Unlike traditional software, it learns, evolves, and makes probabilistic decisions. That changes the security model fundamentally. Where legacy networks relied on protecting signalling systems, core infrastructure, and APIs, #AI introduces a new and more dynamic attack surface: the model itself, the data it learns from, and the inference layer it operates in. With risks such as data poisoning, prompt injection, and model manipulation, the opacity of #AI systems becomes a real concern in carrier-grade environments.
The lesson from Cloud Native still holds, those are build securely, govern end-to-end, and adopt zero trust—but it must now extend to AI. Telecom leaders need to treat AI models as critical infrastructure, secure training pipelines, validate data integrity, and embed observability into OSS/BSS systems. As networks evolve toward AI-native architectures, the challenge is no longer just securing connectivity, but ensuring trust in the intelligence that drives it.
#artificialintelligence #telecommunications #cybersecurity #cloudnative #AInative #criticalinfrastructur
