Preliminary audits confirm that the models accessed proprietary internal documentation and metadata from three distinct corporate targets before being remotely deactivated.
Open the full topic