The atmosphere remains as hushed as ever. As elegant as ever within the Mudam setting. And the conversations are as polite as they are uncompromising. A welcome luxury at a time when the financial centre must transform itself and must… take on the new risks arising from this transformation, against a backdrop of over-regulation, a complex value chain and an interconnected ecosystem that will undoubtedly mean any problem faced by one player will affect all the others in the financial centre.
“Artificial intelligence and digitalisation will have an impact comparable to that of the 19th-century Industrial Revolution,” he asserts 
Seated around the table, the CEO of HSBC Europe, 
A unique ecosystem… provided the risks are managed
This statement stands in stark contrast to the sector’s usual caution. “We probably can’t win the global AI race,” admits Mr Wagner, “but Luxembourg can build a centre of excellence in AI and digital finance.” The country already has several strengths: a flexible regulatory framework, a regulator considered to be approachable, and a dense financial ecosystem.
For Emanuele Vignoli, it was precisely this combination that prompted HSBC to launch its Orion tokenisation platform from Luxembourg. “The two obvious choices were Hong Kong and the UK,” he explains. But the bank ultimately chose the Grand Duchy thanks to “three key factors”: successive blockchain laws, Luxembourg’s financial ecosystem and the presence of a team dedicated to these matters within the CSSF.
The executive then goes on to detail the ways in which tokenisation is already being used within his group. HSBC has launched tokenised deposits that operate 24 hours a day, 7 days a week for international treasurers. The aim is to enable large companies to move their liquidity instantly, without needlessly tying up large sums over weekends or outside banking hours. “This isn’t a new currency,” he insists. “It’s existing money, but in digital form.”
The technology does not stop at cash flow management. The system also allows for the integration of automated mechanisms similar to smart standing orders, driven by AI. These tools can understand a company’s payment habits, anticipate cash flows and automatically execute certain transactions.
Growing complexity and American dominance
At PayPal Europe, Sean Byrne describes a similar trajectory. The group already makes extensive use of machine learning to detect fraud and analyse billions of transactions. But the arrival of generative AI takes things to a whole new level. “We’ve moved from ‘yes/no’ answers to massive inference capabilities,” he explains. The group has created an executive role dedicated to “transformation and simplification through AI” and is setting up a European centre of excellence bringing together around 100 specialists. PayPal also expects to save $1.5 billion over several years thanks to this transformation.
Beyond the technological demonstrations, concerns quickly emerge. Raoul Mulheims, founder of Finologee, takes on the role of ‘bearer of bad news’. His assessment is blunt: European financial institutions are facing increasing regulatory complexity, particularly with Dora, obligations relating to technology subcontractors and confidentiality constraints.
The main problem, he argues, is that the most advanced AI models are still largely dominated by a handful of US companies. “We are now talking about infrastructure,” he stresses. Yet financial institutions must strike a balance between technological performance, professional secrecy, regulatory compliance and digital sovereignty.
This situation is creating a growing gap between large corporations and medium-sized players. The largest banks can develop in-house infrastructure or negotiate directly with the tech giants. Smaller players, however, risk being left behind with limited tools. The result: the phenomenon of ‘shadow IT’ is on the rise, with employees discreetly using unauthorised AI tools in their work environment.
Biba Homsy also highlights the fragmentation of the regulatory and technological landscape. Stablecoins, tokenised deposits, central bank digital currencies, multiple blockchains, and legal differences between countries: “this isn’t a solution for tomorrow,” she warns. “The obstacles are technological and legal, but also stem from the multitude of players and services.”
Operational resilience: a key issue
A few minutes later, 
However, the focus of his speech is on three topics: tokenisation, the European savings and investment union, and artificial intelligence. On tokenisation, Claude Marx confirms that Luxembourg is in a unique position thanks to its four blockchain laws, its ‘technology-neutral’ approach and its ongoing dialogue with market participants. He acknowledges, however, that projects are progressing more slowly than hoped, particularly due to the historical absence of a European digital settlement currency.
However, the regulator sees an imminent acceleration. “The question is no longer whether the technology is viable,” he says, “but whether we have the talent and the plan needed to be part of this change.”
Claude Marx takes a much more proactive stance on artificial intelligence. In his view, the financial sector has entered a ‘third wave’ of AI, following the initial statistical models and then generative AI. This new phase is based on ‘agent-based’ systems, capable of acting autonomously. He cites, in particular, the recent example of Anthropic’s Claude Mythos model, which is capable of identifying IT vulnerabilities and generating functional exploits without human supervision. “The capabilities of AI now far exceed those of the best human experts,” he asserts.
The head of the CSSF emphasises, above all, how rapidly these models are advancing. “We have technology that is becoming 10 to 100 times better and cheaper every year,” he warns. “What is impossible today could become possible tomorrow. And tomorrow is Thursday or Friday.”
The assessment is particularly damning regarding certain aspects of the financial sector. Claude Marx openly criticises the anti-money laundering systems still in use at many institutions, which can generate up to 90% false positives. “It’s a waste of resources,” he says, whereas AI-driven systems could automate a large proportion of the checks.
The real challenge, however, is neither technological nor regulatory. It concerns skills. “The most worrying aspect is talent and people,” he explains. For the regulator, standard training in AI or prompting is no longer enough. The sector must now “upskill” 10 to 20 per cent of senior executives and middle managers so that they truly understand the evolution of attacks, models and risks. “Companies are organised around two- to five-year horizons,” he concludes. “But that is no longer the pace of the world of exponential AI or quantum computing.”
Despite these warnings, Claude Marx remains convinced that Luxembourg can capitalise on this new technological revolution. He cites the country’s political stability, its triple-A credit rating, its ability to adapt, and the close relationship between the public authorities and the private sector. “Luxembourg has always been agile,” he sums up. “And I am convinced that we will remain competitive.”
