Breaking news

EBA Moves To Simplify Banking Regulation With New One-Step Approach

Harmonised Retail Diversification Framework

The European Banking Authority (EBA) has issued its final guidelines on proportionate retail diversification methods under the Capital Requirements Regulation. These guidelines provide a unified framework for assessing retail portfolio diversification, ensuring that smaller institutions can benefit from a proportionate regulatory approach.

Enhanced Proportionality For Smaller Institutions

Under the guidelines, institutions seeking the preferential 75% risk weight for retail exposures must demonstrate sufficient portfolio granularity. Individual exposures to a counterparty or a group of connected clients should not exceed 0.2% of the total eligible retail portfolio. The framework introduces flexibility for smaller institutions, allowing them to qualify for the preferential risk weight if no more than 10% of the portfolio exceeds the 0.2% threshold.

Simplified Regulatory Procedures

The EBA initially considered two assessment methods: an iterative baseline approach and a one-step alternative. In the final version, the Authority adopted the one-step approach to simplify implementation and reduce operational complexity. The diversification threshold was also raised from 5% to 10%, reflecting feedback from financial institutions and aimed at reducing disproportionate regulatory pressure on small and medium-sized banks.

Clarified Treatment Of Securitised Retail Exposures

The guidelines also address the treatment of securitised retail exposures, distinctly outlining the criteria for institutions acting as originators versus those acting as investors. For investor institutions, a temporary, limited derogation has been introduced for cases where obligor-level information is unavailable. In such circumstances, the diversification condition may be regarded as fulfilled, thereby permitting the application of the preferential risk treatment despite the lack of detailed data.

Copyright Law Struggles To Keep Up With AI Training

Courts Are Still Applying Old Copyright Rules To AI

AI companies train models on enormous amounts of published material, including books, articles and academic research. Whether using that content without authors’ permission violates copyright law remains unresolved.

Much of the debate centres on fair use, which allows copyrighted material to be used without permission in certain circumstances. Courts consider factors such as the purpose of the use, how much material was involved and its impact on the original market.

Anthropic Case Sets An Important Precedent

A major case involving Anthropic and a group of authors provided one of the clearest rulings so far. Judge William Alsup found that using copyrighted books to train AI models was lawful, comparing the process to people reading and studying literature before creating something new.

Anthropic was nevertheless ordered to pay $1.5 billion in a settlement. The penalty concerned books the company had obtained from illegal online libraries rather than the AI training itself.

For AI companies, that distinction could prove significant because it separates studying copyrighted material from directly copying it.

Competition Could Be The Key Issue

A case involving Thomson Reuters and Ross Intelligence offers a different perspective. A court ruled that Ross could not claim fair use after using Reuters’ copyrighted material to develop a competing AI-powered legal research platform.

The decision suggests courts may be less willing to consider AI training fair use when copyrighted content is used to build a product that directly competes with the original.

For authors, an unresolved question is whether AI-generated content should be considered competition for the works used to train these models.

The Law Has Yet To Catch Up

US copyright law predates generative AI by decades, leaving courts to apply old principles to new technology. Questions also remain over copyright protection for AI-generated works. In Thaler v. Perlmutter, a court ruled that material created entirely by AI cannot receive copyright protection.

Major AI companies remain involved in copyright litigation, and different courts could reach different conclusions. For now, there is no universal rule: the legality of AI training will depend on the circumstances of each case and how courts ultimately interpret copyright and fair use.

Aretilaw firm
eCredo
The Future Forbes Realty Global Properties
Uol

Become a Speaker

Become a Speaker

Become a Partner

Subscribe for our weekly newsletter