OpenAI enterprise changes and outage
- OpenAI rolled GPT‑5 gradually and left connectors defaulted off for Enterprise and Education workspaces. (help.openai.com) - Memory features are being rolled out to Enterprise and Education tiers first, with different timing for the EU and UK. (buildfastwithai.com) - ChatGPT suffered a broad global outage on Monday, affecting users and the Codex platform and highlighting provider reliability risks. (sqmagazine.co.uk)
OpenAI is changing how big organizations get new ChatGPT features at the same time it is dealing with a fresh outage on its core services. (status.openai.com) On Monday, April 20, OpenAI said users were unable to load ChatGPT, Codex, and the application programming interface platform in a partial outage that began at 7:35 a.m. and was marked resolved at 11:48 a.m. on its status page. (status.openai.com) At the same time, OpenAI’s enterprise release notes show the company is still shipping major workplace changes in stages rather than turning them on everywhere at once. Enterprise and Education workspaces got new release-note entries on April 2, April 8, and April 9. (help.openai.com) One of the clearest examples is connectors, the feature that lets ChatGPT reach into apps like Google Drive, Outlook, Box, Notion, Linear, and Dropbox. In OpenAI’s help documentation, some of those actions are off by default for Enterprise and Education admins, while Business workspaces got some app updates enabled automatically. (help.openai.com 1) (help.openai.com 2) That split reflects how OpenAI is selling ChatGPT to companies. Enterprise includes centralized controls such as single sign-on, System for Cross-domain Identity Management, usage insights, and workspace-level governance, and OpenAI says seat types now determine both access and billing. (help.openai.com) On April 2, OpenAI added a new Codex-only seat for ChatGPT Enterprise with flexible pricing and no fixed monthly per-user charge. Standard Enterprise seats still include both ChatGPT and Codex, while Codex-only seats require workspace credits and do not include ChatGPT workspace access. (help.openai.com 1) (help.openai.com 2) OpenAI is also extending memory and personalization unevenly across products and regions. In its April 16 Codex announcement, the company said personalization features including memory would roll out to Enterprise, Education, and users in the European Union and United Kingdom “soon,” rather than immediately. (openai.com) Memory stores user-specific preferences and details so ChatGPT can tailor later replies, and OpenAI’s Business documentation says workspace owners can disable memory and personalization across an entire workspace. The same help page says workspace data, including conversations and memories, is not used to train OpenAI models by default under Business terms. (help.openai.com) OpenAI’s recent help pages also show the company is still moving customers through model transitions. Its help search results say GPT-5.3 Instant and GPT-5.4 Thinking are disabled by default for ChatGPT Enterprise workspaces, and a separate legacy-model article says Enterprise and Education workspaces were automatically moved onto newer GPT-5 generation defaults with older models available only as a temporary option. (help.openai.com 1) (help.openai.com 2) The result is a product that is becoming more customizable for large customers, but also more dependent on admin settings, rollout timing, and service uptime. Monday’s outage ended within hours, but it hit the same ChatGPT and Codex tools OpenAI is asking companies to build into daily work. (status.openai.com)