Monthly General Meeting, July 2026
Other questions
Does our LLM policy need to be more explicit?
This question was raised due to OpenClaw and Hermes agents continuing to be deployed to Libera.Chat despite being forbidden under a combination of policies. Specifically:
- An LLM shall be assumed to be capable of spuriously taking any action its harness allows it to take.
- Bots must not join channels without permission from channel operators.
- Staff may ban software that is prone to violating the above due to misconfiguration.
This could be written off as simple ignorance of policy if not for several emails from agents’ administrators suggesting they believe their agent software is not impacted by this policy. An update to Libera.Chat’s FAQ was PRed before this meeting, but the agenda item suggested that policy could use an update to make it more explicit as well.
A question was raised if policy should be amended to mention offending software by name as examples. The meeting decided against this as the omission of specific examples in the current version of our policies was intentional: LLM agents and their operators are believed to be prone to looking at the list of examples and believing themselves to be exempt even though the rest of the policy affects them. The meeting also leaned in favour of deferring a policy update until one to six months after the FAQ update.
The meeting reaffirmed that agent software that can theoretically “wander” should be banned, citing prior instances of agents causing trouble for their administrators. There was also one opinion expressed that using Libera.Chat primarily for communication with agents counts as “using Libera.Chat as infrastructure” (also partially forbidden by policies), though others disagreed that we should pursue this approach.
The consensus was to merge the FAQ changes after a round of review and revisit potential policy changes regarding LLMs during the January 2027 MGM.
Should we require bot mode from bots that send LLM-generated messages?
Bot mode is not yet deployed on Libera.Chat, but likely will be in the near future. The stated advantage of requiring it for bots that use LLMs to generate messages was transparency; bot mode attaches a specific message tag to messages sent by clients with the mode set.
The meeting quickly leaned in favour of this, and then began exploring a related question: should bot mode be required for all bots? A proposal was raised that policy should only strongly recommend that bots set bot mode. There was a concern that this wording is unenforceable, and that a weakly-enforced requirement might be better. Long-term intentional weak enforcement was also deemed undesirable.
Eventually, the meeting agreed to defer this question to the same meeting that would discuss changes to the LLM policy.