Mace AI is funding a doctoral student in UVA’s Aikyam Lab to help advance research into how AI systems make decisions—and how those systems can be understood and evaluated as they take on more responsibility.
Led by Assistant Professor Chirag Agarwal, the lab studies AI interpretability, explainability, safety and trustworthiness. The fellowship supports research, student training and the development of methods and tools for evaluating advanced agentic AI systems. It covers the 2026–27 academic year, with an option for renewal.
“The support from Mace AI helps further expand our work and creates opportunities for students and researchers to contribute to some of the most important challenges facing AI today,” said Agarwal.
Why support research outside the company?
For Mace AI co-founder and Chief Scientist Greg Frank, the value of university research goes beyond access to new ideas. It gives researchers a setting in which to develop methods that others can inspect and independently verify.
In a conversation with James Heckman on TheStreet Roundtable, Greg explains why Mace AI chose to support this work at UVA rather than keep it entirely in-house.
“Safety should belong to everybody.”
He points to the university’s research environment as a place for creative minds to work together, while emphasizing that methods for assessing AI safety should be open to scrutiny beyond the company developing them.
Watch the conversation
Understanding the decisions behind the actions
As organizations give AI agents access to sensitive systems, understanding how those agents make decisions becomes increasingly important.
“Companies and universities are giving AI agents privileged permissions faster than we can explain how those agents are making decisions,” said Stephen Turner, assistant dean for research at UVA’s School of Data Science. “This interpretability gap is where Chirag and the Aikyam Lab come in.”
Greg raises a related research question in the interview: how should harm be measured when it is distributed across several agents working together? Examining each agent individually may not explain the effect of the group. Answering that question requires methods other researchers can examine and challenge.
The fellowship supports the time and academic work needed to investigate questions like these. For Mace AI, that is an investment in research the wider community can learn from—not simply another claim about what our own technology can do.
Read more
TheStreet covers Mace AI’s fellowship at UVA’s School of Data Science and its support for independent AI safety research.
Read the full article on TheStreet →