Same flaw, opposite verdict: what counts as a vulnerability in AI agents?

A recent article discusses the issue of what constitutes a security vulnerability in AI agents, highlighting that the same flaw can be viewed differently. The article explores the discrepancy in verdicts on vulnerabilities in AI agents. It is available on Medium, with a corresponding discussion on Hacker News. The post has garnered some attention, with a point assigned to it.