JD Vance claimed this week that Iran submitted three different versions of a 10-point diplomatic proposal during nuclear negotiations, and that one of them was written by ChatGPT.
The accusation is notable less for the geopolitical drama and more for what it reveals about how AI tools are showing up in places most people wouldn't expect. ChatGPT drafting a diplomatic document isn't technically surprising - the tool is genuinely good at producing polished, structured text in any format you ask for. What's unusual is that it apparently made it into an official submission without enough editing to hide the seams, or that someone decided the AI draft was good enough to send alongside two human-written versions.
Vance didn't specify which version was the AI-generated one or how the U.S. detected it. Text generated by large language models (AI systems trained on massive amounts of written content to predict and produce human-like text) doesn't carry a reliable signature. There's no watermark baked into ChatGPT's output, and AI detection tools have well-documented accuracy problems - they generate false positives on human writing and miss plenty of AI-generated content. So whatever method the U.S. used to identify the document, it wasn't a simple scan.
For practitioners watching AI's spread into professional and institutional work, this is a data point rather than a scandal. Governments, law firms, and negotiating teams have been using AI writing tools since 2023. The question was never whether it would happen in high-stakes contexts, but when we'd hear about it publicly. The answer, apparently, is now.