From 6G to fraud detection: Three ways AI is redefining telecom

Artificial intelligence (AI) is unsurprisingly front and centre of conversations about the future of telecom networks. In a recent discussion with Mobile World Live, Frédéric Leclercq, Chief Technology Officer, Mobile Connectivity Solutions, discussed how AI is already shaping the way networks are built, operated and secured – and why its rise makes cybersecurity and post-quantum cryptography (PQC) more important than ever. 

This article recaps the key insights from that discussion, highlighting three areas where AI is beginning to make a tangible impact across the telecom ecosystem.

AI-native network architectures will define the move toward 6G

At the network architecture level, the next few years will see not just a switch towards AI-enhanced operations, but the very infrastructure itself designed around it.

The switchover to 6G will provide an opportunity for architectures to build AI from inception at every layer to provide real-time inference, training and automatic network provisioning and optimisation for different services. 

AI models will run on network equipment at the edge, rather than in centralised cloud environments – allowing for faster decision making. The vast training inputs of network telemetry data will help AI models rapidly learn, again without the centralisation – and risks – that come from having to aggregate data in one place. 

For operators, this marks a shift towards more predictive, autonomous networks capable of adapting dynamically to changing conditions and service requirements. 

Faster fraud detection and AI-driven telecom security

Another significant trend around AI for telcos will be within security and it’s a rapidly evolving area. This ongoing change within the AI ecosystem is the leading AI-related security concern, with 70% of respondents ranking it among the top three sources of risk in the 2026 Thales Data Threat Report. 

Alongside cybersecurity for AI itself comes the prospect of using AI for cybersecurity operations. We may well see zero-day exploit detection become a reality, for example, with AI being able to identify unusual code execution patterns in network elements before exploits are publicly known.  Recent developments from Anthropic point in this direction: its unreleased “Mythos” model reportedly identified thousands of previously unknown vulnerabilities across major operating systems and browsers, including the ability to autonomously discover and exploit zero-day flaws.

When it comes to network activity, there will be gains to be made around fraud detection thanks to AI. Traditional rule-based systems flag suspicious patterns based on predefined rules, while AI can recognise normal behaviour baselines and detect subtle anomalies. 

For detecting cyberattacks, automated threat notification is already real, but orchestrating response and containment is yet to come. AI will one day be able to isolate compromised endpoints from traffic, rotating credentials and route around attacks automatically.
Operators must train teams to work alongside AI agents, establishing governance for autonomous responses and building Security Operations Centres (SOCs) that have AI at their core, not as a bolt-on.

It’s here the reality that AI and PQC are interdependent becomes clearest. If these AI-powered security systems rely on vulnerable cryptography, adversaries with quantum computers could potentially manipulate or undermine AI-driven defences. This makes it only more important for telco operators to have quantum-safe networks ready. 

At Thales, we are already demonstrating how this transition can work in practice. In 2023, Thales showcased the world’s first PQC-secured phone call, using its Cryptosmart application integrated into a 5G SIM and running on commercial smartphones over a live 5G network. 

Thales has also partnered with SK Telecom to demonstrate quantum-resistant cryptography on a commercial 5G standalone network, showing that post-quantum security can be deployed in real-world telecom environments. 

Intelligent connectivity management through AI

The third trend will be around AI’s ability to make connectivity more personalised, context-aware and ultimately invisible. Depending on the application, location, cost and user preference, devices will automatically connect to the best network for the use case at hand. 

As connectivity becomes critical for more objects, from vehicles and medical devices through to industrial equipment and robots, AI can help optimise what network to use, depending on the environment, data plan or quality of service needs.

For example, an eSIM may connect automatically to the most cost-effective network in each country as someone travels, based on their past usage patterns, while IoT and connected devices could use different connection types, in line with how they’re being used. 

Protecting data and all its sources in an AI-driven telecom ecosystem

At Thales, protecting data at creation, in motion, in use and at rest has been our role for decades, and now we’re taking that experience and applying it to the opportunities and challenges presented by AI. 

Thales AI Security Fabric delivers the first runtime security capabilities designed to protect Agentic AI, LLM-powered applications, enterprise data, and identities. 

As telcos scale their AI adoption, they need assurance that sensitive data, applications and user interactions will be safeguarded – and AI Security Fabric offers an expanding array of tools to secure agentic AI data access, and ensure unified, compliant management of interactions between users, models and data sources.

Across multiple fronts, telcos must prepare for an AI-driven future – across their operations, their management and throughout it all, their cyber resilience. An insecure AI system is worse than no AI – it’s a vulnerability.