Co-design of LLM-based preference agents: participation may drive overtrust

This feed contains recent research on LLM governance, reliability, safety, compliance, human reliance, accessibility, and organizational impacts. Key security-relevant themes include opaque and unstable model behavior across deployment configurations, overtrust caused by participatory AI design, copyright-infringing agent actions, gaps in AI Act compliance validation, and dependency on LLMs for decisions and work. No direct exploitation of software vulnerabilities or actionable cyberattacks is described.