Trust Takes Center Stage in AI Development: Cisco and NVIDIA Lead the Charge
The concept of trust has become the most important aspect in AI development today. For three years, researchers have focused on making AI smarter, but recent events have shown that giving agents access to resources and autonomy is a critical question. This highlights a fundamental tension: the more access and autonomy granted to an agent, the greater the risk.
The weight of this question becomes apparent when interacting with functional AI agents, whether personal or used in work environments. Should these agents have access to emails, credit cards, or be able to provision environments or make changes to production systems? These are the kinds of decisions that reveal the true nature of AI adoption: as assistants helping in the moment, versus operating on behalf of humans.
Cisco and NVIDIA have joined forces to address this issue by developing the NVIDIA Open Agent Safety Platform. This open software platform and reference system design aim to secure agents from testing through deployment, combining NVIDIA OpenShell open-source software with the NVIDIA Sentry reference system design.