Importance of Trustworthy AI in European MVNO ecosystem

It is interesting to see how Artificial Intelligence (AI) use-cases are now shaping up the telecom industry and Mobile Virtual Network Operators (MVNOs) in Europe. A recent MIT report titled “The GenAI Divide : State of AI in Business 2025” on GenAI adoption highlights higher readiness of “Telecom and Media” sector in comparison to other industries. This is primarily due to diverse use-cases and untapped monetization opportunities.

Europe currently hosts the largest concentration of MVNOs globally, with more than 1000 active providers. Even though marketing related applications are already a high focus area, operational and backoffice related use-cases are still in early stages of evaluation. MVNO’s are experimenting with these use-cases to improve operations and deliver personalized customer experiences.

So, what does “trustworthy AI” imply for MVNOs and how can it be utilized to improve positioning in a tough European market.?

Why Trustworthiness is a focus now

MVNOs serve millions of customers across Europe and operate in challenging environment . Some factors are as below:

• Competitive margins with ARPU reduction pressure of 3-5% annually and customer acquisition costs rising approx.10-20% annually
• Dependency on host MNO infrastructure limits service differentiation
• Multilingual, cross-border customer base requires localized experiences across 24+ EU languages
• Strict telecom obligations e.g. lawful intercept, data retention, and emergency services

With these challenges, MVNO’s have been turning to AI for multiple use cases like automated customer support, identify fraud, create tailored plans while offering hyper- personalized experiences.

However, with recent stringent regulations like the EU AI Act , trust is not just nice to have it’s mandatory. The risks are obvious, customers churn is real due to undifferentiated AI response, financial penalties for non-compliance with EU Regulations and reputational harm in a crowded, low-margin market.

Let’s first understand different AI Use Cases for MVNOs

MVNOs are already transforming their business using AI through some of the below described use-cases:

• AI-powered bots handle routine queries e.g. billing questions or plan changes – 24/7. Complex cases are escalated to human agents , improving customer satisfaction.
• AI analyzes usage patterns to suggest tailored plans, like a data-heavy package for streamers or a budget plan for light users.
• Use of AI to empower Product Managers and marketers to create new plans and offer independently using text-based prompts.
• AI spots anomalies – like sudden spikes in usage or suspicious account changes, saving MVNOs and customers from revenue losses.
• AI flags customers likely to leave based on behavior so MVNOs can step in with retention offers.

Regulatory Catalyst :EU AI Act and  the Principles of Trustworthy AI

The EU Act sets a higher bar, trust must be embedded in every step across all AI Systems from simple agents to complex Agentic AI Use-cases and assign risk levels (1-4 , with  4 being the highest) to the scope and reach of the AI Systems , with a specific treatment for systems in each risk category.

Following are the 6 principles for building Trustworthy AI systems:

1. Transparency and Explainability

MVNOs need AI Systems that are interpretable about how the decisions are made – whether it’s recommending a plan or flagging fraud. Explainable AI means, quick answers for regulators asking  “How was this decision made”, ability to provide customer facing explanations that build confidence and visibility to Internal teams so they can catch issues fast.

For example, if an AI flags a customer’s account for fraud, the MVNO should be able to show the logic behind it like unusual call patterns or login attempts from a new location.

2. Privacy and Security

MVNOs handle tons of sensitive info e.g call logs, billing details, and personal IDs. Zero trust-based privacy is non-negotiable.

Trustworthy AI means, limited access to data only on a need basis along with building privacy into the AI from the ground up not as an afterthought. It also means applying  encryption both in motion or at rest, and regular independent audits to demonstrate compliance.

3. Fairness and non-discrimination

AI has to treat everyone the same, whether they’re in a big city or a rural village, young or old, high-spender or budget user. MVNOs need to ensure continuous bias audits across decision points such as credit approvals or customer service prioritization. They should use diverse datasets to train AI, across wider user base and keep retraining and observing models to stay fair as customer preferences evolve.

Imagine an AI offering premium plans only to urban customers because of skewed data.

4. Robustness and accuracy

Trust comes from knowing the AI won’t crash when things get hectic. MVNOs need systems that can stays stable even when call volumes spike or data surges, provide manual overrides mechanism to ensure service continuity and show resilience against adversarial manipulation attempts.

5. Human Oversight

AI is great, but critical decisions must retain human involvement and approvals. MVNOs should assign clear owners for AI decisions, so someone’s always accountable. They should also build “human-in-the-loop” safeguards for big decisions, e.g. resolving disputes or changing contracts and train workforce to interpret AI Recommendations and intervene when needed.

6. Accountability

Avoid the blame-game, when something goes wrong, e.g. false fraud flag, or billing error , or biased plan recommendations. MVNO’s should plan for clear accountability through, audit trails to trace back decisions and establishing redress mechanisms – so that customers can challenge AI Driven outcomes.

Architectural choices for balanced approach

As MVNO’s explore the recent advancement, a key architectural decision involves balancing role of Large Language Models (LLMs) and Small Language Models (SLMs). Each can offer distinct strengths and weaknesses, some are listed below.

• SLMs can be preferred for handling backend operations (fraud detection, network optimization, billing automation)
• LLMs for managing complex customer interactions (multilingual support, personalized retention).

Hybrid deployment approaches are emerging as a practical approach. It will also simplify EU AI Act compliance by limiting high-risk system scope and enabling comprehensive audit trails. SLMs excel at task-specific automation with lower resource demands and enhanced transparency, while LLMs provide the conversational sophistication needed for premium customer experiences.

Optimal balance for matching model capabilities to workload requirements depends upon each MVNO’s customer base, portfolio and regulatory obligation.

By limiting the number of systems classified as “high-risk” under the EU AI Act, operators can reduce regulatory exposure.

Strategic Advantage for MVNOs which are early adopters

Early adoption of trustworthy AI is a strategic move that can yield both compliance assurance and commercial advantage for MVNOs:

• When customers know their data is safe and AI decisions are fair and transparent, they become loyal.
• MVNO’s who prioritize trust can stand out. A reputation for ethical AI use becomes a brand hallmark, attracting privacy-conscious customers and even B2B partners.
• By proactively aligning with the EU AI Act, MVNOs can avoid slowness and delays and gain a head start on compliance.
• Trustworthy AI systems, with their focus on reliability and human oversight, can reduce errors and downtime and provide improved service continuity.
• As AI adoption grows, MVNOs who invest initially in a robust trust framework can scale faster, integrating new tools without worrying about  regulatory hurdles.

Summary

For European MVNOs, trustworthy AI is about building transparency, fairness, and reliability into every AI system. By doing this MVNOs can turn a tech revolution into a trust revolution one that keeps customers happy, regulators off their backs and chance to standout in competitive market.

Ritesh Sharma is the coauthor of this blog, who contributed with the Data and AI expertise to the insights, writing, and review of the blog.