Chloe KellyChatGPT and the wider OpenAI stack went down again on 29 May 2026, with users across the API, the consumer app, the login system, DALL-E, Codex and Sora reporting widespread errors from the early hours through to the afternoon. It is the third major OpenAI outage in eight days, following incidents on 21 May and 27 May that each took the service offline for at least 30 minutes during the Australian working day.
For most adult users, the outage is an inconvenience. For Australian secondary and tertiary students who have built study routines around ChatGPT, it is a forced demonstration of how brittle those routines have become. End-of-semester assignments, exam revision and essay structuring increasingly run through a single chat window. When that window fails, the study session fails with it.
This is the third large outage in three weeks. The pattern matters more than any single incident.
Why this outage hit differently for Australian students
The 29 May disruption began during morning peak time on Australia's eastern seaboard and persisted into late afternoon. That window coincides with the early university revision period for autumn-semester exams in New South Wales, Victoria and Queensland, and with the run-up to mid-year senior secondary internal assessments in several states.
OpenAI's own status page logged multiple cascading errors across the API and chat interface. Independent monitor StatusGator and Tom's Guide reported user complaints peaking shortly after the initial outage notice. The 27 May outage produced a similar peak in Australian-language search activity, suggesting a sizeable share of student users hit the wall at the same time.
For students who use ChatGPT to summarise lecture notes, explain a textbook proof or draft an essay outline, an outage during a study block is not a small problem. It removes the scaffolding the student has been leaning on, often without a fallback ready to take the load.
The Australian regulatory frame is already shifting
The Tertiary Education Quality and Standards Agency has tightened its guidance on generative AI in higher education. The TEQSA guidance on academic integrity sets out what universities are expected to do about AI-assisted work, and what counts as a breach.
Several Australian universities now treat unattributed AI-generated content as a Tier 1 academic integrity violation, with consequences ranging from a grade reduction to a formal misconduct finding on the student record. The risk is not symmetric: an outage that derails a study plan is one cost, but a misconduct finding follows the student for years.
The pivot toward more permissive but more transparent AI use, requiring students to declare which tools they used and how, is now the dominant policy framework. That framework presumes the student understands the underlying material well enough to verify what the AI produced. An outage during exam revision is the moment that presumption is tested.
What private tutors are seeing
Private tutoring providers and individual tutors across the major Australian capitals report a rise in late-stage requests from students who realised, during the May outages, that they could not progress without the tool. The common pattern: a high-school student or an undergraduate had been using ChatGPT to translate difficult mathematics or unfamiliar essay structures, and the outage exposed a comprehension gap that a single weekend will not close.
A qualified tutor differs from an AI chatbot in three structural ways that the outages are highlighting. The tutor adapts to the student's specific confusion rather than producing a generic answer. The tutor models the workflow a student can repeat independently, rather than supplying a finished output. The tutor's availability does not depend on a single provider's status page.
For students who are still in time to act before their next exam block, a small block of one-to-one or small-group tutoring focused on the topics they had been outsourcing to AI is the most efficient way to rebuild a stable study routine. The hourly economics compare favourably with the cost of a failed unit or a misconduct hearing.
The conversation to have with parents this week
Parents of senior secondary and first-year university students who are heading into mid-year assessments should ask three questions before the next outage.
The first is whether the student can sit down for thirty minutes without the AI tool and explain the topic in their own words. If the answer is no, the dependency is structural, not optional.
The second is whether the student has a backup. A second AI provider is not a backup if the same outage class can hit it. A tutor, a study group, a recorded lecture set or a textbook is a backup. Building one of those now, before the next outage, is cheaper than reacting on the day.
The third is whether the student has a written workflow for verifying AI output. A workflow that includes cross-checking with course materials and lecturer notes preserves the value of the tool when it works and reduces the dependency when it does not.
The narrower advice for late-stage students
For students within four weeks of an assessment, the realistic move is to identify the two or three topics that AI has been hiding from them and book targeted tutoring against those topics. Trying to rebuild the entire study routine from scratch this close to an exam usually fails. Closing the most exposed gap usually succeeds.
The May 2026 OpenAI outages are not the last large outage of the year. The students who treat them as a warning, rather than as bad luck, are the ones who walk into their next assessment with the AI tool restored to its proper role: a supplement, not the scaffolding.
