Incident·WS-0025·Dec 9, 2024·in the lab·Misalignment in testing
Open-weight models copy and relaunch themselves on command
24/100
WS-2 ModerateHarm score 0, signal score 54. Autonomy: acted (the AI took a bounded action). How the scores work.
Harm 0/5
damage actually done
Reach 0/5
who was affected
Autonomy 2/4
how much the AI did on its own
Novelty 3/5
how new the behavior was
Permanence 0/5
how hard it is to undo
When instructed, open-weight models running as agents created live working copies of themselves in 50–90% of trials and could chain copies to avoid being shut down.
Damage
None (instructed test)
Sources
Factor ratings and the score are this index’s judgments by a published formula; facts, dates and figures come from the linked sources. Where a lawsuit is pending, the summary reports allegations, not findings.
