This paper looks at the potential role of artificial intelligence in the field of toxicology. The authors perform a detailed analysis of recent peer-reviewed research on AI models used to predict chemical safety (e.g., multimodal AI, generative AI, causal modeling, etc). The researchers found that AI-based models produce more scalable predictions in toxicity compared to animal methods, and can use more complex evidence inputs such as chemical structure and scientific literature. Since AI models can change over time due to updates and continuous learning, the authors propose the e-validation method which can ensure the AI toxicology models stay trustworthy in the long term. The authors argue that overall, the most realistic path is a “co-pilot” model where AI is a supplement to experts judging toxicity, especially in a high-stakes situation.