Olly - August 30, 2026 13:35

A couple of recent posts have claimed that I fed years of a customer's personal writing into an LLM without their consent.I want to clarify what actually happened. I used Claude Code to help write a parser for a WordPress export file, which I then used to import the blog into a Pagecord account. Note that I use Claude to assist most of Pagecord's development, and I've been open about that for years.I didn't submit a customer's writing to a model. Claude Code did have access to the export file while I was working on the parser, so yes, some of it was read by an LLM. But the work was on the parser, not the writing: it reads the XML file, wrangles some WordPress HTML, then saves the resulting posts to the Pagecord blog. I didn't treat the file as sensitive at the time because it contained public blog posts, but of course public doesn't mean the customer agreed to it being read by an AI coding tool. I should have clarified what I planned to do – I'm sorry I didn't. I'll make sure I do so…