ChatGPT Agent Security: What Dots Change

ChatGPT Dots are OpenAI’s always-on agents, launched in September 2026, that operate on dedicated cloud computers, retain context, use connected applications, and can continue assigned work between conversations, shifting agent security concerns from supervised tasks to persistent digital identities. Available to eligible Pro and Business Premium users and as an admin-enabled beta for Enterprise, Edu, and Healthcare workspaces, Dots can reportedly connect to more than 4,000 apps and use controls such as Custom Rules, automated reviews, mandatory hand-offs for selected sensitive actions, and monitoring. The text emphasizes that these safeguards do not eliminate indirect prompt injection, excessive permissions, persistent-memory abuse, or data-exfiltration risks, particularly when agents have credentials, broad app access, schedules, and external communication capabilities. It also identifies beta limitations including no data or inference residency, exclusions for FedRAMP and EKM workspaces, incomplete audit visibility, limited control over individual dot memories, and cloud environments that may not inherit organizational device or network policies. It recommends treating each dot like a service account by restricting permissions, using dedicated read-only accounts where possible, keeping local computer access disabled, testing for prompt injection, maintaining external logs and agent inventories, and limiting deployment to small pilots, especially for regulated or sensitive workloads.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.