Over 70% of Users Reluctant to Grant AI Agents Full PC Control, Survey Shows
A poll conducted by Android Authority revealed that over 70% of respondents are unwilling to grant autonomous AI tools like OpenClaw full access to their personal computers. Specifically, 43% cited security concerns as their reason for refusal, while 27% reported having no interest in the capability altogether. As AI shifts from simple conversational chatbots toward autonomous agents that directly manipulate operating systems, privacy boundaries and control mechanisms have become key bottlenecks for wider adoption. Without robust permission isolation and safeguards against system exploitation, user reluctance will continue to hinder the adoption of computer-use AI tools. Around 25% of respondents were open to giving AI system-level access provided proper safety safeguards are implemented, while less than 4% were unconcerned about the risks. The survey highlights risks like prompt injection attacks, where malicious actors can manipulate the AI to execute unauthorized actions and compromise the user's host environment.
## BACKGROUND
OpenClaw is an open-source autonomous AI assistant designed to execute system tasks locally—such as managing emails and calendars—via chat platforms like WhatsApp and Telegram. Prompt injection is a major security vulnerability in large language models (LLMs) where carefully crafted inputs trick the model into overriding its safety rules and executing malicious commands.