OpenAI tightens enterprise controls

OpenAI is tightening the default gates on workplace data inside ChatGPT while it slowly expands access to its newest flagship model. (help.openai.com 1) (help.openai.com 2) In OpenAI’s ChatGPT Enterprise and Education release notes, updated Box, Notion, Linear, and Dropbox apps are disabled by default, and workspace admins or owners must turn them on in settings. OpenAI says those updates add new app actions, including write capabilities where the connected service supports them. (help.openai.com) OpenAI’s help documentation says Enterprise and Education workspaces also keep some unified Google Drive actions off by default until an admin enables them, while ChatGPT Business keeps those Google Drive actions on by default. The same documentation says connectors were renamed “apps” on December 17, 2025, as OpenAI merged connected tools into one interface. (help.openai.com 1) (help.openai.com 2) At the same time, OpenAI’s release notes say GPT-5 is “slowly rolling out” across ChatGPT Plus, Pro, Team, and Free on web, mobile, and desktop. The company says GPT-5 will reach Enterprise and Education “soon,” but did not give a date in the notes published last week. (help.openai.com 1) (help.openai.com 2) That split shows two tracks inside OpenAI’s product strategy in April 2026: broader access to core models for general users, and tighter admin control over how workplace data enters the system for managed customers. OpenAI’s apps documentation says admins can review each app and its actions before enabling them for a workspace. (help.openai.com) (help.openai.com) The underlying product issue is simple: these apps let ChatGPT pull from services such as Google Drive, Box, Notion, and Dropbox, and in some cases write back into them. For an information-technology team, that turns a chat tool into a bridge to internal files, tickets, and documents, so default settings become a security and governance decision. (help.openai.com) (help.openai.com) Futurum Group cast OpenAI’s newer GPT-5.3 Instant Mini in a different role from a headline model launch. In an April 12 analysis, Futurum said the model is aimed at fast, lower-cost, latency-sensitive enterprise work rather than maximum capability, and cited its survey showing 67 percent of organizations already run generative artificial intelligence in production and 75 percent plan to increase budgets. (futurumgroup.com) OpenAI’s own model release notes show the company has been tuning that speed tier too: on March 16, 2026, it updated GPT-5.3 Instant to improve follow-up tone and reduce teaser-style phrasing. That places model rollout, admin permissions, and app-level controls on the same timeline rather than as separate product stories. (help.openai.com) For enterprise buyers, the near-term question is less whether ChatGPT can connect to company systems than who gets to switch those connections on. OpenAI’s latest notes answer that with a default of admin approval first, model access second. (help.openai.com) (help.openai.com)