Banks must find ways around the ‘black box’ problem that affects many artificial intelligence-driven models, the Federal Reserve’s Lael Brainard says.

The Fed governor told a virtual conference hosted by the central bank that supervisors would expect financial firms to be able to explain how their models are reaching certain results. Without this, it would be impossible to tell if a model was producing biased results, she said.

Because AI algorithms leave the computer to fit a model to the