ChatGPT Dots updates and changelog (OpenAI Dots) since the 29 Sep 2026 launch through 16 Oct 2026
Dated changelog of ChatGPT Dots (OpenAI Dots) since the 29 September 2026 launch. Each entry states what changed and what it means for day-to-day use, checked through 16 October 2026.
This page is a dated changelog for ChatGPT Dots (also called OpenAI Dots) since the 29 September 2026 launch. Each entry states what OpenAI changed and what that means for you day to day.
Sources checked through 16 October 2026: the ChatGPT Learn note for 5–9 October, OpenAI Help (Getting started with your dot, Dots privacy, security, and safety FAQs, Manage dots in ChatGPT workspaces), the ChatGPT release notes, and Meet dots.
What changed after launch
14 October 2026 — Troubleshooting answers in Getting started, and GPT-5.5 retires
What changed: Help's Getting started article now ends with troubleshooting answers. GPT-5.5 leaves ChatGPT, ChatGPT Work and Codex on all plans on 14 October 2026.
What it means for you:
- Each task has its own conversation. A correction you give in the main Dot chat does not automatically reach a delegated task, so put the result you need (and any files or links) in that task's own thread.
- Changing the selected computer does not move an existing task. Look for work on the computer or cloud environment where the task started. Local tasks show in the sidebar on the connected computer; recurring work is under Scheduled.
- Read access on a connected app does not always include write or send. An expired connection comes back as a reconnection request.
- Sign-ins on your personal computer do not carry into the Dot's cloud browser.
- Local computer access is granted only from the ChatGPT desktop app on that computer. Connecting the computer to Codex or turning on Work Sync is not the same step. If Allow access is missing, you are not in the desktop app on that machine.
- On mobile, plugins, custom rules and memory sit under Customize in your Dot's profile; opening your Dot's computer puts you in control.
- At launch your Dot still cannot call you or have its own standalone email address. Texting remains a limited beta for some US Pro users.
- Update saved model settings, scheduled tasks and scripts that still select GPT-5.5. Dots run on GPT-6 Astra, so the Dot itself is not retired with GPT-5.5, but a Work or Codex schedule it depends on may still name GPT-5.5.
More detail on these controls sits on the features page and in the FAQ.
13 October 2026 — Same 5–9 October fixes shared again
What changed: Community and social posts repeated the 5–9 October Learn note. ChatGPT Learn's What's new list still pointed at that note as the latest Dots update, so treat those posts as a re-share, not a separate release. Several fixes from the note are worth calling out on their own:
What it means for you:
- Shorter waits for replies, smoother transcript scrolling, and a fix for excess network use.
- No extra phone notifications while you are already in ChatGPT.
- Correct character display in Safari and Firefox on the web.
- Setup can autofill email verification codes; Outlook setup stuck states are fixed.
- More stable safety banners on the web; Enterprise access issues on the web and when starting a voice call are fixed.
5–9 October 2026 — Create a Dot on your phone; clearer Codex thread choices
What changed: OpenAI's 5–9 October note lets you create a Dot in the ChatGPT mobile app and improves how Dots reuse Codex context.
What it means for you:
- On your phone you can name the Dot, customize its look, connect plugins, and open ChatGPT straight into that Dot's conversation. Get started lists the mobile app, desktop app and desktop browser as equal starting points. Mobile web is still unsupported; connecting your computer still happens in the desktop app.
- Your Dot can search across your ChatGPT conversations, use context from existing Codex threads and automations, and is better at continuing an existing Codex thread versus starting a new one. It can also read and edit ChatGPT Work automations.
- Threads it manages can rename themselves as the work changes, with fewer redundant "created task" messages and faster sidebar loading.
- Failed activity loads can be retried; approval buttons are no longer clipped by oversized Gmail previews; Outlook setup getting stuck is fixed.
First-hand reports of missing sidebar tasks or tasks created outside the intended project are collected on Dots and Codex.
9 October 2026 — Privacy FAQ on memories and one-off approvals
What changed: The privacy FAQ clarifies what you cannot delete piece by piece and how proactive research and approvals work.
What it means for you:
- There is no per-memory delete. You cannot view, delete or edit individual Dot memories (including details that arrived through plugins). Clearing the Dot's own context means deleting the Dot.
- Disconnecting an app or turning off ChatGPT Memory stops new sharing only; what the Dot already has stays.
- Proactive research tools are read-only: they cannot message other people, change content through plugins, or control a browser or computer. Follow-up actions still go through normal approval rules.
- One approval covers one action. Approving a message does not grant ongoing permission to contact people. For a future send, name who, what, and when or under which conditions.
- A Dot can have its own separate Slack account. Card purchases on a merchant site need your approval; advance approval is only valid when it specifically covers that purchase.
8 October 2026 — Your own computer and Codex cloud in Getting started
What changed: Getting started now covers optional local computer access and Codex cloud environments.
What it means for you:
- Access to your own computer starts off. You connect it from the ChatGPT desktop app on that computer and confirm Allow access; Revoke access turns it off. With it on, your Dot can create Work or Codex tasks, use local skills, and use your local browser when the cloud browser is blocked. The desktop app is available on Windows.
- Codex cloud tasks need an environment you create in Codex first.
- Custom rules have four behaviors: Take action without asking, Take action if pre-approved, Ask before taking action, and Hand off to you. "Pre-approved" means you asked for that action explicitly in your prompt. Learn describes the same control in its own words, so labels may not match Help word for word.
- Pause lives in the ••• menu on the Dot's profile. Reset deletes the Dot along with its conversations, saved memories and scheduled tasks.
- You can connect only one personal computer at a time, and that computer must be online with the ChatGPT app open.
7 October 2026 — Pro 500; GPT-6 in Chat leaves the Dot model alone
What changed: Release notes introduce Pro 500 and reopen Pro 200. GPT-6 with Intelligent UI arrives in the Chat tab.
What it means for you:
- Pro 500 is $500/month with the highest included Pro usage and Astra Ultrafast in ChatGPT Work and Codex. Among Pro plans, Ultrafast comes only with Pro 500; buying credits on Pro 100 or Pro 200 does not unlock it. Pro 200 is open again at $200/month with changing included allowances.
- Learn lists Dots access as Pro 100, Pro 200 and Pro 500 for users over 18 outside the EEA, the UK and Switzerland. Ultrafast is described for Work and Codex; none of the pages says Pro 500 makes the Dot itself faster. Help only notes that later you may scale each Dot's speed or monthly workload.
- Work and Codex models are stated as unchanged in that batch. Dots remain powered by GPT-6 Astra, so this release does not announce a new Dot model.
About 4 October 2026 — Enterprise workspace admin guide
What changed: Manage dots in ChatGPT workspaces documents admin permissions and cloud computer controls for Enterprise.
What it means for you (admins and members):
- Four Dots permissions: Use dots (Beta), Add dots to Slack and Microsoft Teams, Allow local computer access, Use custom rules for dots; plus three cloud computer controls and a separate password manager switch.
- Workspace model settings do not apply to Dots.
- Only the Dot's owner can direct it in Slack. Other people's DMs or mentions do not start work; channel members still see what it posts.
- Phone messaging through iMessage, RCS or WhatsApp is not available for Enterprise at launch.
Where the official pages still disagree
First-month usage
- Release notes: for the next month, Dots usage does not count toward eligible Pro, Business and Enterprise plan allowances; usage terms per plan come later.
- Help: your plan includes an allowance for deeper work, with extended limits for the first month after launch.
- Learn: conversations with your Dot do not count toward ChatGPT usage limits; tasks your Dot starts or manages in Work or Codex count toward those products' limits as usual.
Read together: chatting with your Dot is the low-risk part. Work and Codex tasks it launches can still draw on limits, and later terms are not published yet. If the month counts from 29 September, the window ends around 29 October. Plan detail lives on pricing and usage.
Enterprise wording
Help calls Enterprise access (including Edu and Healthcare) a beta an admin must turn on. Learn says Enterprise is rolling out worldwide and off by default. Both agree nothing happens until an admin enables it.
Texting and calls
Help still describes texting as a limited beta for some US Pro accounts through a third-party provider, not available in Business or Enterprise. Learn's overview omits texting and lists ChatGPT, calls, Slack and Teams. On calls the pages fit: you can call your Dot with the phone button in its conversation; your Dot cannot start a call to you at launch.
What Pause actually stops
Help shows Pause and Resume as one switch. Learn is more precise: Pause stops the Dot's current main task; delegated tasks stop in Activity; recurring tasks cancel in Scheduled; ending a voice call does not necessarily stop assigned work. Stopping a task does not undo what it already did.
Launch baseline (29 September 2026)
At DevDay OpenAI introduced Dots as always-on agents powered by GPT-6 Astra, each with its own cloud computer. They went to Pro outside the EEA, UK and Switzerland, Business Premium in all supported regions, and Enterprise as an admin-enabled beta, rolling out gradually. You create a Dot on the desktop app or desktop web, not on mobile web (mobile app create arrived later — see 5–9 October). Launch channels were ChatGPT, Slack and Teams, with texting marked as coming soon. For Europe, see Dots in the EU and UK.
Copyable asks based on these changes
These asks are written by gptdot.net from the official wording above; they have not been run as tests on this site.
Before you pause or step away:
List everything you are doing for me right now in three groups: your main task, delegated tasks in Activity, and anything in Scheduled. For each one, tell me whether Pause would stop it. Don't stop anything yet.
To see which work can draw on limits during the first month:
For the work you did for me this week, tell me which parts were plain conversation with you and which ran as Work or Codex tasks. List the Work and Codex tasks with a one-line description each.
Because you cannot delete single memories:
Don't save anything about my health, money or family from my connected apps into your own notes unless I ask you to. If you're unsure whether something counts, ask me first.
To make a future send unambiguous:
Next Friday at 9 AM my time, send the weekly update to the two addresses I give you below, using the draft we agree on today. If anything in the draft changes before then, ask me again before sending.
Related
- Pricing and usage: plan allowances and what counts toward limits
- Access and login: where to create your Dot and missing-entry checks
- Dots not showing?: rollout delays and current error reports
- What is ChatGPT Dots?: approvals in practice
- Dots in the EU and UK: the regional rules in full
Copy this task brief
A dated log of official Dots changes since launch, where Help, Learn and the release notes disagree, and four copyable asks based on those changes.Download Markdown ↓