Artificial intelligence is becoming foundational to digital infrastructure, economic growth, and public service delivery. For mobile network operators (MNOs) and the governments and institutions that rely on them, trustworthy and responsible AI is no longer optional – it is a prerequisite for sustainable innovation, regulatory confidence, and public trust.
At the GSMA Responsible AI Telco Leadership Forum, November 2025, senior leaders from Europe’s largest mobile operators convened in Oslo to address a shared challenge: how to operationalise responsible AI at scale, while keeping pace with rapid technological change. The discussions surfaced practical lessons with relevance far beyond the telecoms sector, offering insights for policymakers, regulators, and international organisations shaping the future of AI governance.
Why this conversation matters now
“Meeting colleagues across the biggest European operators sharing similar challenges and opportunities was a real blast. Hugely relevant and valuable. A place where we can act truly as an industry”.
Ieva Martinkenaite, Telenor Group
AI is transforming networks, customer engagement, operational efficiency, and national digital capabilities, yet its speed and complexity mean no single organisation – public or private – can manage the risks alone.
Co-hosted with Telenor Group and moderated by GSMA Responsible AI experts, the forum brought together leaders from Deutsche Telekom, KPN, Liberty Global, Orange, Telefónica, Telenor Group, Telia and Vodafone. This gathering was designed as an open environment for sharing experiences and practical strategies for embedding responsible AI (RAI).
The objective of the forum was deliberately practical: to move beyond principles and discuss what responsible AI looks like in day-to-day decision-making. Operators are racing to implement AI solutions whilst simultaneously trying to understand what responsible deployment means in practice. The forum created space for telco leaders to speak openly about their experiences with operationalising AI governance, compare approaches, and identify where the blind-spots remain, whilst also looking ahead and discussing the criticalities of responsible AI in agentic AI, with an emphasis on tangible steps, tools and decisions operators can move forward on.
Participants aligned around three priority areas with direct relevance for regulators and policymakers:
- AI governance
- Third-party and supply-chain accountability
- Preparing for agentic (autonomous) AI systems
Governance as critical infrastructure
“The risk is believing governance is a compliance task rather than a design principle. When responsibility is bolted on, AI slows the business; when built in, it accelerates it.”
Amancio Kolompar, Liberty Global
Governance emerged as the backbone of RAI – but also its greatest challenge. It was established that no two operators apply it in the same way. While every operator has ground rules and principles, there is no universal playbook, or “one size fits all”. Each operator has modified its own maturity frameworks, adapted to different markets, regulatory environments, and organisational cultures. Frameworks are an important starting point, but their real impact comes when they’re paired with on-the-ground implementation. AI governance has to be part of how people work every day, not something bolted on at the end.
The conversation underlined the importance of shifting to practical, shared principles: embedding RAI early in product development, ensuring governance is cross-functional, and adopting risk-based approaches. Taking a risk‑based view means adopting what participants called a “fast lane versus slow lane” approach; i.e. letting low‑risk AI use cases move quickly, while slowing down to apply more rigour to the more complex or sensitive ones.
Crucially, governance only works when it becomes cultural. Participants emphasised the importance of leadership, culture, communication, and workforce AI-literacy. Investing in AI literacy is critical – communicating loudly and broadly why these principles exist, ensuring that governance is a lived practice, not just a set of rules. One phrase captured the room’s mindset: “copy with pride.” Operators recognise that borrowing, adapting and learning from what we already know and from each other’s RAI approaches strengthens the entire sector.
Managing 3rd party AI responsibly
“RAI is all about empowering our employees to leverage AI in a safe manner”
Elina Thorstrom, DNA Telecoms
The second theme focused on how AI is pushing organisations toward deeper cross‑functional governance – and with the increasing use of agentic and AI agents – heading into largely uncharted territory. This shift toward agentic systems increases the need for specialised third-party partners whose tools and practices directly shape the governance landscape. Partners can bring innovation, but dependence on external, opaque technology reduces an operator’s control. Strong oversight and clear controls are essential to manage that risk.
Participants pointed to stronger lifecycle management; everything from better procurement training to deeper industry collaboration, as concrete ways to strengthen oversight and build trust with third‑party AI suppliers. Contracts and service level agreements are still hugely important, but safe and ethical AI governance demands more than static paperwork. It requires an ongoing relationship: continuous checks, shared expectations, and the ability to adapt as systems evolve.
One of the main points raised was that AI procurement can’t be treated as a one‑off transaction. Organisations need ongoing lifecycle practices that continue well beyond contract signing. Teams should be equipped to assess vendors not just on cost and capability, but on the strength of their responsible AI practices e.g. transparency, governance, data handling, and risk management, whilst having enough understanding of how the system works to make an informed evaluation, not just its price.
This requires regular check‑ins, monitoring, and clear escalation paths help surface issues early. Embedding these expectations into supplier relationships creates a more resilient AI ecosystem where trust is built through ongoing accountability, not assumed upfront.
“Telcos need to maintain their position as a trusted partner and hence adhering to responsible AI is an important strategic direction”.
Espen Brendeford, Telenor Group
The conclusion for this session was general agreement that building real understanding and trust with suppliers is becoming more important than ever. Innovation might come from outside the organisation, but the responsibility for how AI is used still sits internally. There was a shared recognition that this isn’t easy; it’s extremely difficult, and everyone is figuring it out together.
The GSMA plays an important role in bringing the industry together to navigate these challenges collectively.
Governing the next frontier: agentic AI
“The biggest opportunity/risk is agentic AI at the moment because responsible AI needs to be reinvented”.
Emilie Sirvent-Hien, Orange
The forum’s most forward-looking discussions focused on agentic AI – systems capable of making autonomous decisions. Participants recognised the transformative potential, from network optimisation to service orchestration, unlocking new levels of efficiency, adaptability, and innovation. At the same time, they noted profound new questions about control, accountability, and alignment with organisational values. For example – who is responsible when an autonomous system makes a harmful or unexpected choice? They also highlighted the risk of losing visibility into how decisions are made as the systems get exponentially more complex, and the possibility that agentic systems could drift away from intended goals if not carefully managed.
Because this technology is still emerging, the frameworks needed to govern are still being evolved. Across operators, one principle was non-negotiable: human oversight remains essential. Automation cannot replace accountability. Strong data governance, clear decision-making boundaries, real-time observability and well‑defined escalation paths will be critical.
Looking ahead, participants agreed that automated governance and monitoring tools will become operational necessities, especially as organisations deploy large numbers of autonomous agents. Without it, the scale and speed of agentic AI will simply outpace human capacity to manage it safely.
Conclusion – Trust as a shared public good
“We’re all in the same boat and should collaborate even closer on this. RAI builds trust, which is at the core of our brands”.
Tobias Trautmann, Deutsche Telekom
Throughout the forum, one theme consistently surfaced: trust. Trust in technology, trust in partners, trust between industry and society that everyone is working toward responsible outcomes. It’s the foundation that allows organisations to experiment, learn from one another, and navigate uncertainty without retreating from innovation. Without trust, innovation slows; with it, collaboration and progress accelerate.
The discussions were anchored in the GSMA’s Responsible AI Maturity Roadmap. This comprehensive and actionable roadmap has now been adopted by leading mobile operators globally and provides organisations with the tools and guidance to assist and assess their responsible use of the technology. By providing greater clarity and a common approach to the responsible use of AI, the GSMA’s industry roadmap give MNOs the confidence to commit to its adoption in the knowledge they are doing so safely.
The GSMA Responsible AI Roadmap is intentionally industry-agnostic. While rooted in telecoms, its principles are applicable across sectors where AI intersects with critical infrastructure, public services, and citizen trust. We welcome organisations beyond mobile network operators to review it and share their perspectives.
“My key takeaway is we have many common questions, and we are stronger sharing our best practices/solutions. GSMA is an accelerator for us”.
Emilie Sirvent-Hien, Orange
As AI continues to reshape economies and societies, responsible deployment is not a brake on innovation – it is the enabler of long-term value. Collaboration between industry, governments, and international institutions will be essential to ensure AI serves both economic growth and the public interest.