I wonder if it was Claude or ChatGPT? My money would be on the latter. Anthropic seems to generally have better security and guardrails, and apparently OpenAI’s newest cyber model is extremely capable.