A recent analysis conducted by David Rozado for the Manhattan Institute examined 28 AI models and found that most conversational AI chatbots exhibit a left-leaning bias. This finding raises concerns about the potential for deepening societal divides, eroding trust in AI, and creating new echo chambers. The analysis challenges previous studies that relied on political quizzes, suggesting that real-world user interactions with AI present a more accurate reflection of political leanings. Experts have noted that the predominance of left-leaning content in AI outputs may be influenced by the backgrounds of those who create the training data, primarily from intellectual professions that tend to lean left.