BBVA explores new approaches to AI agent security with NVIDIA

As AI agents become more autonomous and take on increasingly complex tasks, companies need to establish clear boundaries for their actions and maintain control over their operations. The newly announced NVIDIA OpenShell Reference Architecture is designed to address these challenges through mechanisms that govern agent behavior and intervene when established security boundaries are exceeded. The news builds on initiatives for AI safety using open-source technology, already bringing together companies across the technology stack, including BBVA.

These capabilities are particularly relevant in sectors such as financial services, where traceability and risk management are essential. For BBVA, the collaboration with NVIDIA provides an opportunity to explore new approaches to embedding security and governance into AI agents from the design stage, in the context of its strategy to industrialize the development and deployment of these systems across the organization to improve customer experience and streamline internal processes.

“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,” said Valentín Sánchez, Global CSO at BBVA. “Working with the technology industry and partners such as NVIDIA gives us access to leading-edge approaches to AI agent security and governance, helping us continue to make progress in this area.”

“For banks to put their proprietary intelligence to work with agents, clear boundaries and oversight need to accompany every step toward greater autonomy,” said Aser Blanco, Global Banking Lead at NVIDIA. “Through our collaboration, BBVA will explore how to enforce those boundaries as agents operate, supporting the bank’s ambition to improve customer experience and employee productivity while maintaining control.”

AI agents can perform a sequence of actions to complete a task, from retrieving information to using external tools. This means that ensuring their security requires more than just controls built in at the design stage; it also depends on the ability to monitor their decisions and actions during execution.

To address this need for continuous oversight, the OpenShell Reference Architecture combines infrastructure for running AI agents and managing their interactions with other tools, with independent security systems capable of detecting and containing actions that exceed established boundaries. The OpenShell secure runtime software is open source and available on NVIDIA’s website and on GitHub.