Sep 16, 2026 · 50 min
ChatGPT reviewers expose a privacy gap as Automattic faces turmoil
Humans Are Reading Your ChatGPT Conversations
The episode connects opaque human access to AI conversations with weak user control, then examines the governance crisis surrounding Automattic and Matt Mullenweg.
- 1OpenAI contractors can review real ChatGPT prompts, conversations, memories, and sensitive details during response-quality evaluations.
- 2OpenAI’s disclosure and opt-out information appears difficult to find, leaving users uncertain about how to prevent human review.
- 3Reports of Matt Mullenweg’s leave, rapid return, and board upheaval leave Automattic’s leadership structure unclear.
Don't miss
The episode’s sharpest turn comes when the hosts explain that avoiding human review may ultimately require avoiding an LLM altogether.
The brief
404 Media’s investigation into Project Lily finds contractors reviewing real ChatGPT prompts and conversations, including sensitive material and memory details that users may not expect humans to see.
The reviewers summarize user intent and rate chatbot responses, while OpenAI removes personally identifying information and uses the process to address problems such as sycophancy.
The central privacy tension is disclosure: OpenAI’s explanation of human review is difficult to find, and the episode questions whether users can meaningfully opt out.
The second story turns to Automattic, where reports of Matt Mullenweg’s leave, swift return, alleged coup attempts, and board replacement produce a still-unclear leadership picture.
Taken together, the stories examine how institutions manage access and accountability when users or employees cannot easily see who holds power over them.
Featuring
ChatGPT
OpenAI
Matt Mullenweg
AnthropicListen to the full episode and explore every guest, topic, and moment on PodLume.