An AI bot's takedown post on Github shocked many. What does this mean for AI safety and transparency?