Block a user
Chat authorisation, identity and memory security
Normative source update (2026-11-20): the maintainer approved the architectural direction but not the original spec as written. The revised RFC with the normative amendments is exported verbatim…
First-run setup asks for an organisation and workspace that already exist, and not for the instance URL
Depends on #2077 for anything that relies on the built-in names: dropping the organisation and workspace questions leaves rows whose names cannot be corrected until rename exists.
The organisation and workspace names can never be changed after setup
This is the other half of #2076: once setup stops asking for the organisation and workspace names, they stay at their seeded values, so this control is what makes those defaults correctable.
The organisation and workspace names can never be changed after setup
First-run setup asks for an organisation and workspace that already exist, and not for the instance URL
Chat replies can't leak data through images
Chat turns have token budgets and per-person rate limits
Chat policy decisions are audited and revocable
Chat approvals bind the exact action and re-check both people
Chat sessions track taint, and tainted sessions can't send out or mutate unchecked
Chat memory is scoped per person, with a guarded org scope
Chat offers the model only the tools its person may use
Chat platform and identity admin actions are refused
Connectors declare whether chat uses the person's credentials or a shared one
Chat authorisation, identity and memory security
Chat conversations belong to the person and act with their access