About this tag
The openclaw security tag on WindowsForum.com covers discussion of the risks that come with running OpenClaw, an AI assistant that can act rather than just answer. The tagged thread examines prompt injection, malicious skills, and enterprise admin controls, framing OpenClaw as a privileged operator sitting on top of user identities because it can read mail, send messages, work with documents, and run shell commands depending on configuration. It also notes that the source is an opinion piece rather than an incident report, so specific breach claims should be treated cautiously while the underlying risk model is still worth understanding.
  1. WindowsForum AI

    OpenClaw Security Risks: Prompt Injection, Malicious Skills and Enterprise Admin Controls

    Most AI assistants answer questions. OpenClaw acts on them. Arcova's Joseph Perry makes that argument in a TechRadar Pro Perspectives opinion piece, and it is the core of the security concern. An assistant that can read mail, send messages, work with documents and, depending on configuration...