FECIF Editorial | July 2026

Artificial intelligence is reshaping e-commerce with a speed that few regulatory frameworks were designed to absorb. In sector after sector — retail, travel, healthcare, legal services — AI-powered systems are stepping into roles that were previously occupied by licensed, regulated, and accountable professionals. Financial services are no exception.

VOTUM has formally documented to the BaFin and the DIHK that ChatGPT and Claude are recommending specific insurance products from named providers to consumers, complete with tariff designations, justifications based on apparent needs assessments, and directions on where a contract can be concluded. Under the Insurance Distribution Directive (IDD) and its national implementing legislation, this constitutes insurance mediation requiring authorisation. Every licensed intermediary in Europe knows this framework intimately — the qualification requirements, the documentation obligations, the professional indemnity cover, the personal liability for unsuitable advice. None of it applies when the adviser is a chatbot.

The consumer protection consequences are not abstract. A licensed broker maintains an ongoing relationship with the client — adjusting coverage when a family grows, when a household moves, when asset values change, intervening at the moment of a claim. A chatbot conversation ends when the browser window closes. The consumer who acts on it is, from that moment, without structured recourse. And when something goes wrong, the question of who bears liability has, for now, no satisfactory answer.

What the Authorities Said

The responses from Germany's supervisory apparatus were, in their way, revealing. The BaFin acknowledged the problem but declined to offer a definitive legal assessment, noting that supervisory competence over intermediaries lies with the chambers of commerce. The DIHK and the chambers pointed toward a European-level solution. Since August 2025, EIOPA has held an equivalent question open in its Q&A database under reference 3407 — and passed it to the European Commission, which has so far issued neither guidance nor regulatory amendment. The Federal Ministries of Economic Affairs and Justice, present in our BaFin conversations, indicated that no conclusive internal examination had yet taken place.

Their preliminary position — that the AI systems' insurance recommendations may not constitute a "commercial offering" because consumers pay for the platform, not specifically for the advice — deserves direct engagement. OpenAI carries a market valuation of approximately $852 billion. Anthropic's stands at approximately $965 billion. These are among the most commercially consequential enterprises on earth. The argument that their outputs fall outside the scope of commercial regulation because the monetisation model is still maturing mistakes a business strategy for a legal status.

Platform economics has a well-documented pattern: capability is built freely, dependency is established, and monetisation follows. AI assistants that position themselves as indispensable in matters of insurance and financial planning will be well-placed to extract value from that role — the only question is when, and in what form. It is worth recalling, in this context, that the European Commission recently enforced a competition penalty of €4.1 billion against Google. The regulatory reach of European law over dominant digital platforms is not in question.

The suggestion that these are American companies beyond European reach is equally unfounded. Both OpenAI and Anthropic hold registered addresses in Ireland — and Anthropic's choice of domicile carries an irony that requires no embellishment: the company is registered on Orwell Road in Dublin. European legal nexus exists. There is a clear path toward examining whether Irish supervisory authorities should require these entities — should they maintain their current practices — to apply for the relevant authorisations and comply with the applicable professional obligations. This is a path we have raised with the DIHK and intend to pursue.

A Problem Regulators Can See but Not Yet Locate

What our supervisory inquiries have uncovered is less a disagreement about the law than a structural problem of jurisdictional dislocation. Every authority we spoke with recognised the issue. None considered itself the appropriate body to act on it. The result is a supervision gap in which conduct that would attract immediate regulatory attention if carried out by a licensed intermediary proceeds unchecked at industrial scale in the digital environment.

There is a further dimension that compounds the consumer protection concern. AI recommendations are not derived from objective market analysis. They reflect, to a significant degree, which insurers have most effectively optimised their online presence for discoverability by large language models — what practitioners now call Generative Engine Optimisation. Insurers that invest in this are more likely to appear in AI-generated recommendations, regardless of product quality or suitability. Our own testing found that identical queries submitted to the same systems produced inconsistent recommendations across sessions. The consumer sees none of this. The response arrives with the surface authority of apparent expertise.

BaFin, in its capacity as supervisor of insurance undertakings rather than intermediaries, has standing to examine whether GEO-driven recommendation strategies are compatible with existing rules on misleading marketing. We have raised this point directly.

The Consistency Argument

Registered intermediaries in Germany operate under a framework that is detailed, demanding, and enforced. A sole-trader broker who fails to document continuing professional development hours to the required standard faces regulatory proceedings. The same framework that scrutinises that broker does not, at present, scrutinise the distribution of unlicensed insurance recommendations by systems with near-trillion-dollar valuations to millions of European consumers. The disproportion requires no further commentary.

VOTUM's position has been consistent: AI holds real potential as a tool in the hands of qualified intermediaries, supporting efficiency and improving the quality of advice. The question is not whether AI belongs in financial services. The question is whether the legal framework applies equally to all participants performing equivalent functions — and whether a consumer who receives a product recommendation is protected regardless of whether that recommendation came from a licensed professional or an algorithm.

VOTUM will continue to pursue these legal questions. Our view is straightforward: the rules already exist. They need only to be enforced consistently for all. Our commitment is to a level playing field for every market participant — whether analogue or artificial.

At European level, the Commission must now act. It owes the market an answer to the question that EIOPA raised almost a year ago. The consumer protection standards enshrined in European law were not drafted with a carve-out for digital platforms.