Artificial intelligence-focused risk and compliance startup Aymara in DC released a matrix outlining different scores for how safe ChatGPT, DeepSeek and their peers really are.