Dear ladies and gentlemen,
It is truly a pleasure to join you all here in Vienna—a city renowned for embracing both tradition and transformation. In many ways, this spirit of balancing the old with the new is precisely the mindset that we, as supervisors, should aspire to.
Innovation in insurance has always been present, with the sector continually changing in response to technological advances, demographic shifts, and financial disruptions. However, the pace, scale, and complexity of technological progress in recent years are unlike anything seen before. The integration of new data sources, sophisticated analytical tools, and a highly digitalised environment is transforming the industry in ways that were unimaginable a decade ago. This transformation has led to much more accurate risk assessments, quicker claims processing, and products that adapt to and are tailored for individual customer needs. For instance, artificial intelligence-powered chatbots now handle customer inquiries, while blockchain technology streamlines policy management and fraud detection. These innovations are reshaping how insurance operates and delivers value to customers.
At the same time, I think we would all agree that the speed and depth of this change raise genuine questions not only for those of us responsible for ensuring the system works fairly and safely, but also for the industry and consumers alike. The question I want to put at the centre today is not whether innovation should happen—clearly, it must. The real issue is how we can ensure that innovation unfolds in a manner that is responsible, resilient, and truly beneficial to people. By "responsible" innovation, I mean not only adhering to regulatory requirements, but also proactively considering ethical implications and the long-term impacts on society. Resilience, in this context, refers to the ability of systems and processes to withstand shocks and adapt to evolving risks, while genuinely beneficial innovation delivers tangible advantages for individuals and communities, such as improved access, transparency, and security. How do we embrace these benefits without losing sight of the protection we owe to consumers?
And for this, let me walk you through three topics today:
First, what is actually happening with Artificial Intelligence, particularly Generative AI, in the European insurance landscape. Second, why data—and new forms of data—matter, and where the real tensions lie. Third, the obligation supervisors have to innovate themselves.
Last year, EIOPA published the results of its first dedicated survey on the use of Generative Artificial Intelligence across the European insurance sector. Market monitoring exercises like this are essential. They give us a clear, evidence-based view of how the market is evolving, and, in this case, they highlight something significant: an industry already moving decisively into innovation.
The survey, which covered a representative sample of insurers across the European Economic Area, shows that 65% of insurance undertakings are already actively using GenAI systems. Most use cases are still at the proof-of-concept stage—meaning that while many insurers are experimenting with GenAI, few have fully integrated these systems into their core business processes, indicating ongoing testing and evaluation before widespread adoption. And let that number sink in: six out of ten insurers are already engaging with this technology in some form. It's clear that GenAI is no longer just a futuristic concept—it's already shaping the industry for a majority of insurers.
Second, this expansion is happening in a controlled manner. Insurers are focusing on back-office applications and "assisted AI systems", which are technologies that operate under human oversight, rather than fully autonomous solutions. Assisted AI systems ensure that human judgement remains central in decision-making processes, providing an extra layer of safety and reliability.
However, the situation is more nuanced. According to the survey, more than a third of firms are already developing customer-facing applications such as chatbots or voice assistants. Additionally, more advanced forms of automation—including "agentic AI", which refers to systems capable of independent decision-making—are expected to grow significantly over the medium term. Agentic AI represents a shift towards technologies that can make decisions without direct human intervention, introducing new possibilities and challenges for the sector.
This is precisely why market monitoring is so important: it allows us to detect trends like this early, while they are still taking shape, and to better understand both the opportunities and the risks as they emerge.
These figures tell us that adoption is happening and that these are not marginal activities. They sit at the very heart of what insurance does: assess risk, price products, handle claims, interact with customers. And while these seem genuinely exciting applications of technology, the same survey also highlights important risks and challenges, and these are precisely the areas where our supervisory attention must remain focused.
Hallucination and model unreliability are among the main risks highlighted by undertakings. Gen AI models can produce plausible sounding but factually incorrect outputs. In insurance, where decisions depend on precise policy wording, this becomes a very real concern. Imagine a customer asking a simple question through a GenAI chatbot about whether their policy covers a specific type of damage. The response comes back instantly, clearly written and reassuring: yes, you’re covered. It sounds authoritative. The customer trusts it. But the answer is wrong. Subsequently, cybersecurity vulnerabilities, issues related to data protection, and challenges regarding explainability also arise.
Adding to the complexity, the survey revealed that the dominant strategy among undertakings when it comes to development of such tools is purchasing off-the-shelf solutions or building on third-party models. This reliance makes effective vendor risk management not just important—it’s critical, as disruptions or changes to these models could have wide-reaching impacts across the industry.
AI is only as powerful as the data that feeds it. And more and more, we are seeing new forms of data being collected, used, and combined in ways that raise genuinely difficult questions. And while the risks highlighted above might be common to the whole financial industry, a closer look at new data sources highlights risks particular to the insurance sector.
Telematics is a good example. Usage-based insurance, where a driver's premium is calculated from real-time behavioural data, is no longer a niche product. And when we move from aggregated, traditional data categories to granular individual data, we move into new territory. The same system that might reward a careful young driver can also identify — and price out — someone who regularly drives late at night because of shift work, or someone whose journey patterns reveal they live in a socioeconomically deprived postcode.
This presents a delicate balance: at its heart, insurance is built on the principle of solidarity, with risks shared across the pool. However, as AI and big data enable increasingly refined segmentation, there is a possibility that individuals may contribute less to supporting one another, resulting in a system that is highly accurate from an actuarial perspective, but one that could pose significant social challenges.
Addressing this concerns and others related to the ethical use of data, EIOPA has been working with a Consultative Expert Group on Data Use, bringing together industry representatives, academics and consumer associations, precisely to address these questions. A report will be published in June this year.
And now I want to turn the lens on ourselves.
Because we cannot talk about overseeing innovation if we are not willing to innovate our own practices. Innovation is not a one-way street.
EIOPA's 2030 Strategy emphasises the use of technology as a key enabler for supervisory simplification, efficiency, and effectiveness. That means using data analytics and AI to identify risks earlier. It means automating routine tasks so supervisors can focus on the things that actually require human judgment. It means standardising reporting formats to make it easier to compare across firms and jurisdictions.
But none of that happens without two foundations.
The first essential component is infrastructure. Robust IT systems, strong data governance frameworks, and efficient operational processes provide the foundation for advanced tools . While these investments may not always attract the spotlight, they are fundamental; without them, aspirations for AI-powered supervision are unlikely to be realised.
EIOPA is actively addressing both. Supervisors, like other stakeholders in the insurance sector, need adequate tools to perform their duties effectively, and this starts with continuous training and knowledge development. Supervisors must build and maintain strong AI expertise through forward-looking resource planning and dedicated centres of expertise. To support this, EIOPA promotes a digital culture and provides training to its members on new technologies, including GenAI, through the Supervisory Digital Finance Academy (SDFA), ad-hoc knowledge exchange platforms, international forum participation, and in-person events such as the Digital Finance Week. In June 2025 EIOPA organised a Pension TechSprint to leverage collaboration with startups and multidisciplinary teams to co-create innovative, AI-enabled pension solutions, demonstrating the value of innovation, cross-collaboration, and technology to close the pension gap.
Equally important is cooperation and coordination. AI is global and cross-sectoral, so supervisors must strengthen domestic and international channels to share knowledge, best practices, and experiences. EIOPA regularly organizes industry roundtables on the AI Act, conducts market surveys like the GenAI survey, and engages closely with the European Commission’s AI Office to clarify sector-specific implications.
Let me conclude by emphasizing that innovation in insurance can be a powerful force for social good. It has the potential to make protection more accessible, more personalized, and more efficient. It can streamline supervision processes and offer consumers greater transparency, choice, and control.
But none of this happens automatically. Real progress requires institutions—both within the industry and in supervision—that are willing to engage seriously with risks and invest in the capabilities to manage them. I am convinced that only through collaboration can we unlock the full potential of digitalization, creating a safer, more innovative, and more resilient insurance industry.
Details
- Publication date
- 15 April 2026