OpenAI turns connectors off by default

OpenAI is rolling GPT‑5 across ChatGPT while switching connectors off by default in Enterprise and Education workspaces. (help.openai.com) On April 9, 2026, OpenAI said GPT‑5.3 Instant Mini was coming to ChatGPT as the fallback model after users hit GPT‑5.3 Instant rate limits. In Enterprise and Education, that fallback is default off and admins must enable it through the Early Model Access setting. (help.openai.com) OpenAI’s help center now says connectors are “apps,” the links that let ChatGPT pull information from services such as Box, Notion, Linear, Dropbox, and Google Drive. In ChatGPT Business, some of those apps are enabled automatically, but in Enterprise and Education, admins are told to turn them on first. (help.openai.com 1) (help.openai.com 2) The split is clearest in OpenAI’s own release notes. Business workspaces get connectors default on, while Enterprise plans keep them default off and route setup through Workspace settings and admin review. (help.openai.com) That puts admin controls at the center of the product rollout. OpenAI’s Enterprise and Education documentation says workspace owners can manage model access, app access, and role-based permissions from central settings before employees use those tools. (help.openai.com 1) (help.openai.com 2) OpenAI is also drawing a line on data handling by plan. Its apps documentation says data accessed from connectors in Business, Enterprise, and Education is not used to train models by default, while consumer plans can allow model improvement if the user setting is on. (help.openai.com) At the same time, OpenAI is still pushing faster models into workplace use. The company says GPT‑5.3 Instant is its default model for logged-in users in ChatGPT, and describes GPT‑5.3 Instant as a fast model for everyday work, learning, technical writing, and translation. (help.openai.com) Futurum Group framed the smaller GPT‑5.3 Instant Mini release as a play for latency-sensitive enterprise workloads, citing its survey of 838 decision-makers in the first half of 2026. The firm said 67 percent of organizations were already running generative artificial intelligence in production and 75 percent planned to increase budgets. (futurumgroup.com) The result is a two-track rollout: faster GPT‑5 variants spread across ChatGPT, while external-system access stays behind admin switches in schools and large companies. OpenAI’s latest release notes show the company treating speed and governance as parallel product decisions, not separate ones. (help.openai.com 1) (help.openai.com 2)