Artificial Pancreas Implantables -- How Healthcare Professionals May Deal With DIY Bio Cases

This batch aggregates 11 arXiv security/ML papers describing new attack surfaces, defensive advances, and standards proposals. Key risks: (1) LLM security — adaptive probe-based steering substantially increases jailbreak success (paper reports harmfulness from 6% to 70%), and a novel optimization-triggered backdoor class can be implanted via numerical side-effects of compilation, achieving ~90% attack success while preserving clean accuracy and evading standard safety tests. (2) Covert egress from LLM agents — a multi-modal covert-channel reference monitor is proposed to detect and mitigate ex