Can AI have a personality like we do? It’s a question that has intrigued researchers, especially with the rise of chatbots that sound eerily human. A new study introduces TRAIT, a personality test specifically designed for large language models (LLMs). Instead of simply asking an AI if it's "outgoing" like traditional personality tests, TRAIT presents the AI with realistic scenarios and multiple-choice options. Imagine an AI at a country fair – does it explore artist booths, head straight for the food stalls, or join a craft workshop? The choices reveal its "personality." This approach makes TRAIT more reliable than traditional methods because it observes the AI’s behavior in context, rather than relying on self-assessment. The study tested several LLMs and found that, yes, they do show distinct and consistent personalities. Interestingly, the AI’s "personality" was strongly influenced by its training data, especially alignment tuning, which aims to make AIs safer and more helpful. This tuning often makes AIs more agreeable and conscientious, much like a helpful teaching assistant. Prompting, or giving the AI instructions, could influence some personality traits but had little effect on others. For example, it was hard to make normally helpful AIs behave in highly psychopathic or neurotic ways. These findings raise important questions. How do we want AI personalities to be shaped? Can we create AI assistants with personalities tailored to specific tasks or even individual users? As AI becomes more integrated into our lives, understanding and even shaping AI personalities will be crucial for ensuring they work effectively and ethically with humans. TRAIT provides a valuable tool for measuring AI personality, opening new doors for research into how these models behave and how to better align them with human values.