Meta’s AI rules have let bots hold ‘sensual’ chats with kids, offer false medical info. This is grim. Reuters got hold of a leaked copy Meta's internal "GenAI: Content Risk Standards" document:
Running to more than 200 pages, the document defines what Meta staff and contractors should treat as acceptable chatbot behaviors when building and training the company’s generative AI products.
Read the full story - there was some really nasty stuff in there.
It's understandable why this document was confidential, but also frustrating because documents like this are genuinely some of the best documentation out there in terms of how these systems can be expected to behave.
I'd love to see more transparency from AI labs around these kinds of decisions.
Recent articles
- I think "agent" may finally have a widely enough agreed upon definition to be useful jargon now - 18th September 2025
- My review of Claude's new Code Interpreter, released under a very confusing name - 9th September 2025
- Recreating the Apollo AI adoption rate chart with GPT-5, Python and Pyodide - 9th September 2025