Claude Code leak reveals a persistent background agent 'KAIROS'

KAIROS is the exact tool you’d want for autonomous killer robots and persistent domestic mass surveillance, the very issues Anthropic just made a lot of noise about not implementing for its government and private contractor clients.

An always-on agent with logs, background workers, scheduled refresh, and webhook/event ingestion is precisely what you need for the supposedly banned activity. On the data-collection end, it can continually ingest reports, summarize communications, watch repositories or analyst workspaces, maintain long-lived memory of persons or entities, etc.

Something like this is likely already in use in Iran and Gaza, as it can continually generate new targets and even pull the trigger to release bombs and missiles.