We express our concerns on the overall benefits of the proposal to implement machine-readable Pillar 3 disclosure across jurisdictions (“this proposal”). Based on discussion with members, we have material concerns about whether the overall effort and cost required to implement this proposal would be proportionate.

 

In just the past few years, frontier models have made major leaps in extracting structured information from unstructured documents, and their usability will continue to increase as AI capabilities mature. Given this trajectory, it is highly probable that AI driven data tool development will outpace the proposed implementation timeline (1 January 2029), and even more advanced capabilities will further reduce the limited value of mandating machine-readable formats for data users.

 

We also consider that the proposed wide-ranging and costly project is disproportionate given users are mainly interested in a small number of templates being machine-readable. Our members observe narrow demand for Pillar 3 reports, so of the 60+ quantitative templates in scope, we anticipate that many would receive little or no use from the public even if made machine readable.

 

Furthermore, more than half of these proposed templates are in “flexible” format, limiting comparability even if they are made machine-readable. Without stronger standardisation, the overall value to users may remain limited.

 

Should the BCBS continue with these proposals, we have set out responses to the consultation questions and provide recommendations.

 

Finally, we outline a set of recommendations for the UK regulator (the PRA ) on how existing UK reporting infrastructures and planned programmes, particularly the Future Banking Data (FBD) programme, could be leveraged to implement the proposed requirements more efficiently.