Central Bank of Uzbekistan launches its own CBU-GPT language model for document processing
The Central Bank of Uzbekistan has integrated the CBU-GPT AI assistant into its digital infrastructure. The tool is designed to automate document processing, handle queries from market participants and citizens, and interpret regulatory requirements.
Details
The new service takes on part of the routine analytical workload for the regulator’s staff. Users can upload files to CBU-GPT to get a structured breakdown on a specific topic.
The model is trained to extract key provisions from lengthy texts and compare old and new document versions. This is particularly useful when updating banking regulations, where changes in capital, compliance, or reporting requirements must be identified quickly.
Beyond legal and financial texts, the assistant can explain complex concepts, organize scattered data, and solve basic programming tasks. The system can also adapt finalized materials for publication across various information platforms.
Country and market
This marks the first time a state supervisory body in Uzbekistan’s financial system has integrated generative artificial intelligence into its daily operations. Until now, the use of LLMs in Central Asia was limited to commercial banks and fintech companies, which primarily used AI for customer support, scoring, and anti-fraud.
The emergence of CBU-GPT highlights the development of SupTech (Supervisory Technology) at the state level. The regulator gains a tool capable of processing a growing stream of data from commercial banks without a proportional increase in inspection staff.
Why it matters
The Central Bank processes hundreds of queries and reports daily that require precise legal assessment. Shifting from entirely manual review to machine analysis reduces response times for financial institutions. Standardizing answers with AI also helps avoid discrepancies in how different departments interpret the same rules.
Launching its own language model shifts the regulator from a conservative agency to a technological market participant, setting a digitalization benchmark for the entire Central Asian public sector.
What’s next
The main practical question remains the solution’s architecture and data security. Using AI in a central bank requires strict control over where uploaded documents containing sensitive financial information are sent.
If CBU-GPT is deployed on the regulator’s local servers without transmitting data to external clouds, it paves the way for deeper model integration into banking supervision processes. In the future, such tools could form the basis for automated interaction, allowing banks to get preliminary assessments of their new products via the regulator’s API.