If ChatGPT would not let you in late on 2 November, you were not imagining the queue. Moneycontrol’s story, timestamped 3 November 2026 at 00:27 IST, says users were reporting trouble with login, the ChatGPT app and Codex, while Downdetector’s complaint line jumped hard toward midnight. I would not call the whole of OpenAI “down” yet. The numbers are real. A company statement is not.
This is the part that matters if you use ChatGPT for college notes, office drafts, or a coding session that was supposed to finish before sleep. A user-report spike is a signal, not a death certificate. Downdetector counts people who say something is broken. It does not, by itself, prove every account in India was locked out. Moneycontrol is careful on that point, and so am I.
What the clock and the graph actually show
According to Downdetector, as cited by Moneycontrol, reports about OpenAI services reached 315 at 11:22 PM on 2 November. The baseline sitting under that spike was 22 reports. Complaints stayed relatively low for much of the day, then rose through the evening, with the graph showing a significant jump close to midnight.
Three hundred and fifteen is not a global blackout on the scale of a national telecom failure. It is also not noise. A baseline of 22 means the late-night pile was roughly fourteen times the usual trickle at that snapshot. If you were refreshing the site around 11:30 PM IST and the page spun, that is the window people were yelling about.
The split is the other number worth keeping. Downdetector’s breakdown, as Moneycontrol printed it, put ChatGPT itself at 81% of reported problems with OpenAI services. The ChatGPT app was 8%. Codex was 7%. Add those and you are at 96% of the complaint pile sitting on the three things most of us actually touch: the chatbot, the phone app, and the coding tool. The leftover slice is small enough that this was not a random grab-bag of OpenAI products.
Login, the app, and Codex are not the same headache
Moneycontrol says users are reporting problems with ChatGPT login, the app and Codex. It also says the outage data does not specify whether those issues are login failures, loading problems, error messages, or some other disruption. That gap is annoying, and it is the whole story if you are trying to decide whether to wait or switch tools.
Login trouble is the classic one. You type the password, the spinner sits there, or you get bounced back to the sign-in screen. On a phone that can look identical to a bad network, especially on a crowded Jio or Airtel evening connection. The app slice at 8% is people who specifically tagged the ChatGPT app rather than the service in general. Codex at 7% is the developer lane: the coding assistant that a lot of Indian engineers now keep open beside VS Code or a terminal, not the chat box they use to rewrite a WhatsApp message.
If your chat tab failed and Codex still answered, you were in a different bucket from someone whose coding session died while the website looked fine. Moneycontrol does not have a clean map of which bucket was larger beyond those three percentages. Do not let a group chat flatten all of it into “ChatGPT is dead.”
What Downdetector can and cannot tell you
Downdetector tracks disruptions from user reports. A rise in complaints can point to a possible outage. It does not independently confirm that OpenAI’s services are unavailable. Moneycontrol states that plainly, and it is the line I would paste into any office Slack before someone declares a vendor outage in a status doc.
The same story says the increase does not establish whether the problems hit all users or only a subset. That is the difference between “my account is stuck” and “the product is down for the country.” A subset outage is still miserable if you are in it. It is also why your friend in the next city might be generating images while you cannot get past the login wall. Geography, account type, app version and which OpenAI surface you opened can all diverge, and this report does not sort them.
I treat a Downdetector spike the way I treat a sudden pile of tweets. If the curve leaves the baseline and stays there, something is off for a lot of people at once. I still want the company, or a second instrument, before I write “global outage” in a headline. Here we have the spike. We do not have the confirmation.
OpenAI has not confirmed this one
Moneycontrol says OpenAI has not been confirmed to have acknowledged the reported problems, or to have explained the spike. As of that 00:27 IST story, there is no quoted status-page note, no timeline, and no “we are investigating” line you can hold them to.
That silence is familiar. Consumer AI tools fail in public and the first hour is almost always user reports, then a status banner, then a short post-mortem that may never name India. Until a banner exists, refreshing login is not a strategy. It is a way to burn the night.
For students who had a deadline, the practical read is simple. If the chat box errors, try once more after a few minutes, then move the draft to a local doc. For developers on Codex, the 7% share means some coding sessions were in the complaint mix, not that every repo in Bengaluru stopped. If Codex is part of a paid workflow, this is a reminder to keep a fallback editor path that does not depend on one remote assistant being up at 11:30 PM.
Who feels it in India
ChatGPT is no longer a novelty tab for a lot of Indian users. It sits inside college assignments, customer-support macros, freelance writing, and late-night coding. A failure that clusters on login and the app hits the phone-first crowd hardest: people who never open a desktop browser and will not patiently read a status page. Codex hits a narrower group, but that group is often on the clock.
Moneycontrol does not give an India-only count. The Downdetector figures it cites are about OpenAI services, not a city-level map of Chennai or Delhi. I will not invent a “mostly India” claim the story does not make. What I can say is that the report landed on an Indian desk just after midnight IST, which is exactly when local users notice a US company’s evening incident overlapping our night.
There is also no workaround printed in the story. No “clear cache,” no “switch to the web from the app,” no region note. The available data does not even say whether the failures were login errors or pages that would not load. Anyone selling you a guaranteed fix tonight is guessing.
How to read the next hour
Watch three things, in this order. First, whether complaint volume falls back toward that baseline of 22 or keeps climbing past 315. A spike that peaks near midnight and fades is a bad hour. A spike that doubles again is a different incident. Second, whether OpenAI actually acknowledges it. Until they do, “down” remains a user-report story. Third, whether the mix stays at roughly 81% ChatGPT, 8% app and 7% Codex. If a new product suddenly dominates the chart, the failure moved.
I would not cancel a subscription over one late-evening graph. I also would not trust a half-loaded chat for anything you cannot afford to redo. Copy your prompt somewhere else before you hit send into a spinning window. If you are on Codex, commit local work before you ask it to rewrite a file you have not saved.
What we still do not know
We do not know the error text people saw. We do not know how many of the 315 reports were from India. We do not know if paid accounts and free accounts diverged. We do not know if the website, the mobile app and Codex failed together or in sequence. Moneycontrol is explicit that the outage data does not specify the failure mode, and that a complaint surge is not independent proof of unavailability.
That is an unsatisfying ending for a trending search, and it is the honest one. At 11:22 PM on 2 November, OpenAI-related reports were at 315 against a baseline of 22. ChatGPT was most of that pile. The app and Codex were smaller but real slices. By 00:27 IST on 3 November, OpenAI still had not been confirmed as acknowledging any of it. If your screen is working again, good. If it is not, you are inside a report spike that has not yet become an official outage.
Comments (0)
Be the first to comment!