BBVA Partners with NVIDIA to Secure Autonomous AI Agents in Banking

BBVA is collaborating with NVIDIA to implement the OpenShell Reference Architecture, a new framework designed to secure and govern autonomous AI agents. For fintech leaders, this move addresses the critical challenge of maintaining oversight as AI systems move from simple information retrieval to executing complex, multi-step financial tasks without direct human intervention.

What was announced

BBVA has announced a collaboration with NVIDIA to explore the OpenShell Reference Architecture, a framework unveiled at the NVIDIA GTC Berlin conference. This partnership is designed to enable new approaches to AI agent safety, security, and governance as BBVA scales the use of autonomous systems across its organization. The OpenShell architecture addresses the challenge of maintaining oversight as AI agents take on increasingly complex tasks, such as retrieving information and interacting with external tools.

The architecture functions by establishing clear boundaries for agent actions and providing mechanisms to intervene when these security limits are exceeded. It combines infrastructure for running agents with independent security systems capable of detecting and containing unauthorized actions in real-time. This continuous oversight is a departure from traditional design-stage controls, focusing instead on monitoring decisions during execution. The OpenShell secure runtime software is open source and available on NVIDIA’s website and GitHub.

For BBVA, the goal is to industrialize the development of these systems to improve customer experience and streamline internal processes. The initiative is particularly relevant in the financial services sector, where traceability and risk management are mandatory. By embedding security into the design stage of AI agents, the bank aims to maintain control over systems that can perform a sequence of actions independently, ensuring that every step toward greater autonomy is accompanied by rigorous oversight.

"We are expanding the use of AI agents across our organization, and ensuring their security is a priority. As these systems become more autonomous, we need to strengthen the mechanisms that allow us to monitor their activities and maintain control,"

Valentín Sánchez, Global CSO at BBVA

The companies involved

BBVA is a global financial services group with a strong presence in Spain, Mexico, and South America. The bank has been at the forefront of digital transformation in banking, often being among the first to trial blockchain and AI technologies at scale. It operates as a major player in the international market, focusing on integrating emerging technologies to modernize its core banking systems and customer-facing applications.

NVIDIA has transitioned from a gaming-focused hardware manufacturer to the primary provider of the infrastructure powering the global AI boom. Its chips and software frameworks are now foundational to the development of large language models and autonomous systems. In the financial sector, NVIDIA’s role has expanded from providing raw compute power to developing reference architectures that address specific regulatory and security needs. The company’s influence in the fintech space is growing as it partners with major banks to solve the "last mile" problems of AI deployment, such as latency, cost, and governance. This collaboration follows NVIDIA's broader strategy of providing the technology stack necessary for enterprise-grade AI safety.

What FF News has reported before

NVIDIA has been aggressively expanding its footprint in the financial sector, as seen when Santander Slashes AI Costs by 99% Using Specialized NVIDIA Infrastructure. This focus on infrastructure is mirrored in the payments space, where Modulr Partners with NVIDIA to Launch Purpose-Built AI Models for Global Payment Operations.

BBVA has also been active in collaborative technological milestones, recently participating in a project where Iberdrola, BBVA, and Santander Launch World’s First End-to-End Blockchain Syndicated Loan. Furthermore, NVIDIA’s commitment to the security side of the industry was highlighted when NVIDIA Commits $250 Million to UK AI Safety and Sovereign Computing Infrastructure.

What this means

The move toward "agentic" AI represents a significant risk shift for the banking industry. While chatbots and predictive models were relatively contained, autonomous agents capable of using external tools introduce a new surface area for operational and regulatory failure. BBVA’s adoption of the OpenShell architecture suggests that the industry is moving away from black-box AI toward a "trust but verify" model where security is decoupled from the agent itself. This puts pressure on other Tier 1 banks to prove they have similar "kill-switch" or monitoring capabilities. The open-source nature of this framework also raises questions about whether standardized safety protocols will eventually become a regulatory requirement for AI in finance.