“Wrong Instructions” Spreadsheet, Impossible to find.

Untangling the Mystery: The Lost Google Spreadsheet of AI Hallucinations and Misbehavior

In the rapidly evolving world of artificial intelligence, staying updated on the latest challenges and anomalies is crucial for researchers and enthusiasts alike. Recently, I encountered a puzzling situation that highlights how difficult it can be to track down critical resources when they vanish from public view.

A while back, around mid-2025, a comprehensive Google Spreadsheet was circulating within the AI research community. This document was reportedly curated in collaboration with a researcher from DeepMind and contained a detailed compilation of recent AI hallucinations and instances of unexpected model behavior. Such a resource would undoubtedly serve as a valuable reference for understanding the pitfalls and limitations of current AI systems.

However, despite extensive searches, I have been unable to locate any trace of this spreadsheet beyond a single screenshot I came across in a YouTube video. The link to this visual evidence can be viewed here: [screenshot link].

This lack of availability raises questions—has the document been removed, or simply hidden from public access? If anyone has information about the status of this spreadsheet or knows where to find it, your insights would be immensely appreciated.

In the fast-moving landscape of AI research, resources like this can illuminate critical issues and steer future developments. Hopefully, it isn’t lost forever, but for now, its whereabouts remain a mystery.

If you’re aware of this document or have leads on similar compilations, please share your knowledge. Let’s keep the conversation alive and continue pushing the boundaries of understanding in AI safety and reliability.