libera.chat, as data

## LLMsLLM usage is *not* being banned outright on Libera.Chat. However,going forward, we are going to require greater transparency fromoperators of bots with LLM features *and* from channels whose operators haveexplicitly permitted LLMs to read or write messages in their channels.First, LLM usage on Libera.Chat must be supervised. If you run a bot thateither uses LLMs to produce message content or uses LLMs to process messages,you must ensure that either you or another human who has agreedto supervise your bot is present in any channels the bot has joined.Second, we’re going to be considering the processing of messages by LLMsto be public logging. This includes cases where messages are submitted toon submitted input, as well as cases where the bot operator asserts thatthe models are being run locally. We have no way of validating either claim,and LLM developers and advocates have an unfortunate track recordof acting in bad faith toward contributors to the common good.This has implications for channel operators. Specifically, channeloperators must inform users that their messages may be processed by LLMsif the operators consented to LLMs processing messages in their channels.In addition, we will also require channel operators to inform users if theymight interact with LLMs, as well as ensure that users can tell whichnicks in a channel correspond to LLM-powered bots.We recommend putting this information in the topic or a `ChanServ` entryAdditionally, we’re going to codify the assumption that an LLM is capableof taking any action permitted by the bot/client whenever promptingoccurs, regardless of what the prompt says. This reflects a growing corpusof real-world examples of AI assistants disobeying prompts and deletingimportant information. One of the practical consequences of this is that**AI assistants, agents, and other LLM-driven autonomous clients will be