Trust is often talked about as the necessary foundation of AI, but in reality, it is not something we can simply declare. It has to be built, tested, and proven over time.
That idea sits at the center of Cisco’s work with the National Institute of Standards and Technology (NIST) Generative AI Program. As AI becomes more embedded in how we work, govern, and connect, the real question is what AI can do and whether we can rely on it when it matters most.
NIST’s GenAI Program takes that challenge head-on by turning trust into something tangible. The program treats trust as a performance standard: something that can be measured, stress-tested, and improved.
One of the most compelling examples of this is the program’s “Cat-and-Mouse” evaluation framework. In this environment, generative AI models create content, while discriminative models attempt to detect whether that content was produced by a human or a machine—and, just as importantly, whether it is credible and accurate. What emerges is a dynamic system that mirrors the real-world tension between creation and verification.
That tension matters. In sectors like energy, water, and government, the outputs of AI systems can shape decisions that impact infrastructure, security, and public trust. The ability to distinguish what is real, what is reliable, and what is safe becomes essential. By simulating these pressures in a controlled but competitive environment, NIST is helping ensure that AI systems are capable and dependable under scrutiny.
At the same time, trust is not only about identifying risk. It is also about consistent performance. The GenAI Code Challenge gets at this directly by evaluating how well AI can generate unit tests for Python code from natural language prompts. At its core, the question is simple: do AI-generated outputs actually work as intended?
Through a global, iterative competition that invites participants from across industry and academia, the program creates a feedback loop where models are continuously tested, benchmarked, and improved in the open. Over time, this process raises the bar for performance, and for confidence in how these systems behave in real-world applications.
For Cisco, participating in this work is a natural extension of how we approach innovation. Taking real-time learnings and applying those insights where and when they matter.
The goal is to ensure that what is proven in evaluation environments translates into how AI is actually designed, secured, and deployed.
This connection between testing and implementation is critical, particularly as the policy landscape around AI continues to evolve. By engaging early with emerging standards and contributing to shared benchmarks, Cisco is proud to help bridge the gap between innovation and accountability—so that the two move forward together.
While NIST is a U.S.-based initiative, the implications of this work are global. The frameworks being developed are designed to scale across borders, offering a common foundation for how AI systems can be evaluated and trusted worldwide.
Ultimately, no one organization can undertake this work alone. It requires continuous testing, transparency, and collaboration across all kinds of sectors and geographies.
Moving trust in AI from aspiration to application requires innovating in a way that people, institutions, and society can rely on. NIST’s Gen AI Program is an important step toward that shared future.
