A study from Common Sense Media reveals that AI tools trained on more data are less safe, raising concerns about the current trend of using massive datasets for AI training. Generative AI products that scrape large amounts of data from common datasets are more likely to generate harmful content, according to the review team's findings. Common Sense's "nutrition labels" for AI ratings highlight Kyron Learning as the safest AI product, while Google's Bard and OpenAI's ChatGPT receive lower scores for privacy.