Ioannis Linaritis (Democritus University of Thrace, Law School; European Banking Institute) has posted Artificial Intelligence in Wealth Management: Fiduciary Responsibility and Regulatory Adaptation under EU Law on SSRN. Here is the abstract:
Artificial intelligence is fundamentally reshaping the wealth-management industry, accelerating its transition from a relationship-based and human-centred professional service to a technologically mediated, data-driven, autonomous form of financial intermediation. Historically grounded in fiduciary-like duties of loyalty, diligence, confidentiality, and acting in the client’s best interests, wealth management has evolved beyond its association with private banking into a holistic service encompassing portfolio management, investment advice, retirement planning, tax and estate planning. Robo-advisors, large language models, and agentic AI systems have democratised access to personalised financial services for retail investors, supporting financial inclusion and capital-market participation. Yet the substitution of human judgment by algorithms raises questions on the viability of the fiduciary paradigm underpinning these relationships.
This article examines the implications of AI deployment in wealth management through the lens of European Union law. It analyses the principal risks associated with AI-enabled advisory and portfolio-management activities, including opacity, explainability deficits, conflicts of interest, discriminatory outcomes, manipulative design, hallucinations, data-quality failures, and threats to market integrity and financial stability. Attention is devoted to the tension between the promises of scalability, efficiency, and personalisation offered by AI and the enduring objectives of investor protection and trust in financial markets. Against this background, a layered interpretation of the European regulatory framework governing wealth-management AI is developed, analysing the interaction between sectoral legislation, particularly MiFID II, IDD, and DORA, and horizontal instruments, including the AI Act, GDPR, and the revised Product Liability Directive. An apparent asymmetry is identified between risk classification and fiduciary intensity, since with the exception of particular components, most wealth-management AI systems are treated as limited-risk despite their potential to affect severely life savings, retirement outcomes, and market stability. It is argued, however, that when interpreted teleologically and systematically, the combined operation of sectoral and cross-sectoral legislation produces a coherent framework of obligations for the AI systems providers and deployers, capable of addressing the distinctive risks posed by AI-enabled wealth management.
A central thesis advanced is that the European Supervisory Authorities (ESMA, EBA, EIOPA), play a decisive role in bridging the gap between technologically neutral statutory provisions and rapidly evolving AI applications. Through guidelines, opinions and other soft-law instruments, these authorities translate professional duties into concrete AI-governance expectations relating to transparency, explainability, documentation, data governance, human oversight, auditability, model validation and operational resilience. Fiduciary responsibility is therefore not displaced but operationalised through an increasingly sophisticated framework of technological governance and supervisory oversight.
The article further situates wealth-management AI within the evolving EU legislative landscape. Recent initiatives, including the CCD II, the DMD II, the Digital Omnibus reforms, and the RIS, reveal a broader regulatory movement towards enhanced rights to human involvement, meaningful explanation, protection against dark digital interfaces, and improved retail-investor empowerment, while also seeking to simplify disproportionate compliance obligations and preserve European competitiveness. Rather than signalling the tightening of regulatory regime for wealth-management AI, these developments point towards the incremental adaptation and refinement of existing legal frameworks.
Finally, the article highlights the challenges arising in non-harmonised components of wealth management, particularly tax advice and estate planning, both closely connected to legal advice. Unlike investment, banking, and insurance activities, these domains remain subject to national professional rules, raising questions regarding professional authorisation, accountability, and the scope of AI-assisted recommendation and decision-making. It is argued that AI systems operating in such contexts cannot lawfully substitute qualified professional judgment and should incorporate effective technical guardrails and human intervention mechanisms to reduce liability risks. The article concludes that AI can be accommodated within the fiduciary architecture of wealth management only through robust governance, proportionate systems-quality management, widespread explainability, effective human oversight, supervisory coordination, and strengthened financial and AI literacy among all relevant stakeholders.
To receive new posts from Legal Theory Blog by email, get a free subscription to Legal Theory Stack.
Lawrence Solum