As user-generated content increases and bad actors become more and more creative, content moderation is becoming a critically important issue for online businesses. Building and improving traditional models for the detection of policy breaches can be time-consuming and resource-intensive, especially when new policies need to be covered, or updates to existing ones need to be made. Re-training models on existing policies with new data can also become laborious. As Head of Machine Learning at Checkstep, I wanted to research how we could use ChatGPT to address these issues. Our research and experiments show mind-blowing results, obtaining better performances in accuracy than state-of-the-art models at a fraction of the cost and in far less time. We want to share our results, and our framework for building content moderation classifiers using ChatGPT, to help other platforms create accurate classifiers, for any text-based policy, without any training data. Topics covered: - How to adapt Generative AI models for content moderation use-cases - Our framework for building effective prompts for content moderation tasks - Our results and comparison with well-known benchmarks