This paper examines the trading behavior of US corporate insiders and certain groups of institutional investors (short‐term, transient, top‐performing, and those with fiduciary responsibility) in the eight quarters leading up to a US firm bankruptcy filing. Using a matched sample based on year, industry, and a probability of future bankruptcy model, we find that US corporate insiders display abnormal reduced net trading activity in the quarters before bankruptcy, with corporate insiders ‘going quiet’ immediately preceding a US bankruptcy. In contrast, we find that our identified types of institutional shareholders display abnormal selling activity several quarters before a US bankruptcy. We then use this information to enhance the bankruptcy‐predictive capabilities of recent machine‐learning techniques such as gradient boosting, as well as the probability of future bankruptcy model. We find that the variables measuring the absolute value of net purchases by US corporate insiders in the two quarters prior to bankruptcy, along with the changes in ownership by specific types of institutional shareholders, improve the out‐of‐sample predictive capabilities of our two different bankruptcy prediction models. Overall, we find that specific types of shareholder display abnormal trading in the quarters preceding US firm bankruptcy, and such information improves the out‐of‐sample accuracy of firm bankruptcy prediction models.