Personal AI agents connecting to enterprise systems raise security concerns
Reported on October 12, 2026, personal AI agents are connecting to enterprise systems through OAuth authorization, passwords saved in browsers, and API keys. Employees do not need to submit tickets, and enterprises find it difficult to register or decommission them. The report states that OpenAI's Dots is limited to enterprise accounts and deployed via Agent 365, xAI's GrokBot is distributed through Cursor, and Meta's Muse is currently limited to the United States and Canada. Most of the new controls added by vendors only cover enterprise-managed deployments; personal agents do not appear in the management console. The author suggests using identity as the control plane to allow safe use rather than banning them outright.