🌍 AI and Telecommunications Trends – Week of September 30, 2025
AI is entering the core of digital infrastructure. Networks are no longer simple data channels: they’re starting to behave like living organisms that think, sense and adapt… but also face new threats. These are five trends showing where we’re heading.
1️⃣ Agentic AI in telecom: networks that act alone
What happened: Current mobile networks still depend on static configurations and human supervision, a model that becomes unsustainable with the growth of traffic and connected devices. Agentic AI proposes nodes and controllers that perceive metrics, reason and act by themselves to maintain service quality.
Why it matters: It transforms operation from reactive to autonomous, reducing failures and costs while improving user experience. Operators can reduce operational costs by 30-40% while improving network availability to 99.99%.
Concrete example: Ericsson has already tested self-healing networks in 5G environments, where agents automatically redistribute traffic in congested cells, reducing dropped calls. These functions are expected to be natively integrated into the RAN Intelligent Controller in 6G.
How it works: AI agents continuously monitor network metrics (latency, throughput, interference), reason about the expected optimal state and execute corrective actions without human intervention. They use reinforcement learning to improve their decisions over time.
Practical implications:
- Fewer interruptions in critical services
- Lower operational costs for operators
- Users with more stable connections even at major events
Source: Ericsson – Agentic AI: Pathway to autonomous networks level 5
2️⃣ ISAC: networks that communicate and sense simultaneously
What happened: With Integrated Sensing and Communication (ISAC), mobile antennas don’t just transmit data: they also capture environmental information (distances, objects, movement patterns). Instead of deploying separate sensors, existing infrastructure functions as a distributed sensing system.
Why it matters: It reduces costs for dedicated sensors and enables urban security services, connected mobility and smart cities. ISAC is expected to reduce urban infrastructure costs by 40-50% by eliminating the need for dedicated sensors.
Concrete example: China Mobile tested ISAC on highways, using 5G signals to detect moving vehicles and send that information to traffic management systems. Although still in standardization phase (3GPP Rel-19), ISAC is expected to be a fundamental part of 6G by 2030.
How it works: The same radiofrequency waves that transmit data are used to measure distances and detect objects through time-of-flight analysis and signal reflections. AI algorithms process these signals to build a dynamic map of the environment in real time.
Practical implications:
- Autonomous vehicles with better environment mapping
- Safer and more connected cities
- More efficient urban infrastructure
Source: TCS – Technology trends 2025: Reshaping the future of telecom
3️⃣ Liquid Neural Networks (LNNs)
What happened: Classic AI models are trained once and degrade in changing environments. Liquid Neural Networks (LNNs), on the other hand, continuously adjust their parameters, functioning like a fluid that responds to context.
Why it matters: They improve network resilience and avoid the need for costly retraining, guaranteeing precision even under unforeseen conditions. LNNs show up to 20% less prediction error in variable environments.
Concrete example: Researchers from MIT and Vodafone have shown that LNNs predict channel quality with 20% less error than traditional models under variable multipath conditions. Today they’re in the lab, but in 6G they’ll be able to run on edge nodes and base stations for live adjustment.
How it works: Unlike traditional neural networks with fixed weights, LNNs dynamically adjust their internal parameters in real time according to context, similar to how a fluid adapts its shape to a container. This allows them to maintain high precision without costly retraining.
Practical implications:
- Less degradation in mobile services
- Dynamic adjustments without human intervention
- Networks more tolerant to interference and weather
Source: arXiv – Liquid Neural Networks: Next-Generation AI for Telecom
4️⃣ Quantum + AI applied to networks
What happened: Spectrum management, routing millions of flows and post-quantum security are problems that classical computing can barely solve. The combination of AI with quantum computing promises to accelerate these calculations and open a new paradigm in telecommunications.
Why it matters: It can solve in seconds problems that today require hours, and offers defenses against quantum threats. Initial simulations show 100-1000x improvements in spectrum optimization and complex routing.
Concrete example: Cisco announced software to interconnect different quantum clouds and offer them as a unified service for network applications. Although in initial phase, these initiatives suggest that future infrastructure will be quantum-classical hybrid.
How it works: Quantum algorithms leverage principles of superposition and entanglement to simultaneously explore millions of possible network configurations, identifying optimal solutions in a fraction of the time required by classical computing. Hybrid AI combines these quantum results with classical learning.
Practical implications:
- Networks more secure against quantum attacks
- Advanced optimization of scarce resources like spectrum
- Fast simulations for urban planning and emergencies
Source: Reuters – Cisco rolls out software aimed at connecting quantum computing clouds
5️⃣ 🚨 AI-powered attacks: the new cybersecurity frontier
What happened: AI also empowers attackers. Adaptive malware, zero-click exploits and hyperrealistic deepfakes mark the new wave of cyber threats.
Why it matters: Traditional security systems were not designed to face attacks generated or powered by AI. A smarter network is also a more attackable network. The global cost of cybercrime could reach $10.5 trillion annually by 2025.
Concrete example: In 2025, the EchoLeak vulnerability enabled a zero-click attack on Microsoft Copilot, capable of extracting sensitive data without the user doing anything. Cases like this show that AI-based exploits are already reality and anticipate a future of automated multimodal attacks.
How it works: AI-powered attacks use generative models to create polymorphic malware that evades detection, deepfakes for sophisticated social engineering, and zero-click exploits that leverage vulnerabilities in natural language processing without user interaction.
Practical implications:
- Need for defensive AI at the same pace as offensive
- Zero-trust architectures across the network
- Greater investment in adversarial testing and live monitoring
Source: Trend Micro – Preventing Zero-Click AI Threats: Insights from EchoLeak
🔗 Conclusion: The digital living organism
The network is evolving like a living organism:
- with a brain (Agentic AI) to think
- senses (ISAC) to perceive
- plasticity (LNNs) to adapt
- and quantum force to solve the impossible
But no organism survives without an immune system. And today, offensive AI is already attacking those same capabilities.
The great challenge will be to ensure these intelligent networks are also resilient, before attackers manipulate them against us.
📚 Recommended readings
AI-Powered Trends Reshaping Telecom in 2025 (Accenture)
Comprehensive analysis of AI-driven technology trends that telcos need to know in 2025.
👉 https://www.accenture.com/us-en/blogs/communications-media/ai-driven-technology-trends-telcos-need-know-2025
5G Americas Highlights the Strategic Role of ISAC for 6G (5G Americas)
Strategic analysis of Integrated Sensing and Communication as a key component for 6G networks.
👉 https://www.5gamericas.org/5g-americas-highlights-the-strategic-role-of-integrated-sensing-and-communication-for-6g/
Operator AI Investment to Exceed $86bn (Juniper Research)
Market forecast revealing massive operator investment in AI infrastructure and capabilities.
👉 https://www.juniperresearch.com/press/operator-ai-investment-to-exceed-86bn/
✍️ Claudio from ViaMind
Dare to imagine, create and transform.
