BNP Paribas has unveiled its proprietary ´LLM as a Service´ platform, offering unified and secure access to large language models across its various entities. Designed and operated by the group´s IT teams, this solution enables internal business lines and functions to easily integrate language models into their workflows through a standardized interface tailored to each unit´s specific needs. The infrastructure, hosted within BNP Paribas data centers and powered by GPU resources, supports open-source models, partner models such as those from Mistral AI, and in the future, models trained on proprietary datasets.
This initiative is a central pillar of BNP Paribas´ technology strategy, which seeks to harness data and Artificial Intelligence to enhance customer personalization and streamline operations. By centralizing resources and ensuring robust security, the platform facilitates lower operational costs and high processing power while promoting consistency in generative Artificial Intelligence adoption. The standardized foundation allows BNP Paribas businesses to focus on developing relevant use cases, with the assurance of data privacy and compliance standards baked in at every level.
The bank is already piloting the platform in diverse operational settings, including use cases such as virtual assistants, automated document generation, and advanced search within documents. Early deployments in subsidiaries like Hello bank! and in internal audit functions have guided iterative improvements, with plans for broader industrial-scale rollout across the group. The platform supports the rapid scaling and deployment of generative Artificial Intelligence projects, streamlining integration into existing processes and enhancing technical and security governance. BNP Paribas views this phased, feedback-driven implementation as a model for secure innovation, balancing local business needs with groupwide standards for efficiency, compliance, and operational excellence.
