Since August 2026, financial companies in Europe have been facing a new layer of regulation related to artificial intelligence. The European AI Regulation (AI Act) now imposes transparency obligations on systems that interact directly with customers, such as banking chatbots or insurance virtual assistants. This change is reshaping how banks, fintechs, and insurers design their digital tools.
Transparency Obligations of the AI Act for Financial Services
Have you ever interacted with a virtual assistant on your bank’s website? Since August 2, 2026, this assistant must clearly inform you that you are conversing with an artificial intelligence. This is one of the transparency obligations set forth by Regulation (EU) 2024/1689, which came into effect on August 1, 2024, with a gradual implementation.
Read also : Discover the latest news and trends that are energizing Breton youth
The timeline breaks down into three concrete phases:
- Since February 2, 2025, prohibited AI practices apply: unconscious manipulation of behaviors, mass social scoring, or real-time facial recognition at scale.
- Since August 2, 2026, transparency obligations apply to all AI systems in contact with customers in the financial sector.
- Starting in December 2027, full obligations for high-risk AI systems will come into effect, covering creditworthiness assessment and pricing in life and health insurance.
To keep up with the latest news on Finance Technique, this type of regulatory timeline is a useful reference, as each phase alters the operational constraints for industry players.
Related reading : Discover the wellness and personal development news not to miss this season
National Supervision of Financial AI: The Case of Germany’s BaFin
A less discussed aspect in the French-speaking press concerns the concrete supervision of financial AI at the national level. In Germany, BaFin (the Federal Financial Supervisory Authority) has taken charge of overseeing AI systems deployed by banks and insurers under the AI Act.
This national authority approach means that each European country adapts supervision to its own financial ecosystem. For example, BaFin checks that automated credit assessment models comply with the documentation and traceability requirements imposed by the regulation.
In France, the question arises as to what exact role the ACPR and the AMF will play in this framework. The European text leaves member states some organizational leeway. For French companies developing AI tools, this institutional uncertainty complicates compliance.
AI Compliance and Risk Management: What Fintechs Need to Anticipate
Fintechs using artificial intelligence for credit scoring or fraud detection are directly affected by the “high-risk” category of the AI Act. Starting in December 2027, they will need to provide detailed technical documentation on how their algorithms function.
In practical terms, this involves three changes in the daily management of these tools:
First, each AI model must be accompanied by a compliance assessment before it is brought to market. This assessment resembles a technical audit: it checks that the system does not discriminate against certain profiles, produces explainable results, and that its training data is documented.
Next, the traceability of automated decisions becomes a legal obligation. If an algorithm denies a loan to a customer, the fintech must be able to explain why, with a level of detail sufficient to satisfy the regulator.
Finally, companies will need to implement a system of continuous monitoring. An AI model that performs well at launch can drift over time, producing biased results as data evolves. The regulation requires regular oversight to detect these drifts.
Tokenization of Financial Assets and Banking Adoption
Alongside the AI regulation, the tokenization of assets continues to progress in the European banking sector. The principle is simple: transform a financial asset (a bond, a fund share, a real estate title) into a digital token recorded on a blockchain.
Why are banks interested in this? Because tokenization allows for reduced settlement times and the fractionalization of assets traditionally reserved for institutional investors. A tokenized real estate fund can be divided into shares accessible from much lower amounts than a traditional investment.
European banks are testing tokenization on bonds and funds, with experiments multiplying over the past few years. The European regulatory framework (notably the DLT pilot regime) offers a legal sandbox for these experiments.
The challenge remains interoperability. Each bank or platform uses its own blockchain infrastructure, and technical standards are not yet harmonized. Without interoperability, a token issued on one platform cannot freely circulate to another, limiting market liquidity.
Digital Tools and Automated Wealth Management
Artificial intelligence is also transforming wealth management. Automated analysis tools now enable faster wealth audits by cross-referencing tax, real estate, and financial data in just a few minutes.
AI does not replace the advisor but changes the distribution of tasks. Projection calculations, tax simulations, and portfolio analysis are automated. The human advisor focuses on interpreting results and client relationships.
This evolution raises a question of responsibility. When an AI tool suggests a wealth strategy that turns out to be unsuitable, who is responsible? The software provider, the advisor who used it, or the financial institution? The AI Act provides some answers by imposing obligations on the system provider, but the chain of responsibility remains to be clarified in practice.
The financial sector is undergoing a period where regulation and innovation are advancing simultaneously. Companies that anticipate the requirements of the AI Act, document their models, and invest in traceability will find themselves better positioned when the obligations of December 2027 come into effect.