ChatGPT users faced service disruptions on September 29 as OpenAI reported elevated errors across ChatGPT, Codex, and its API. The problems affected requests, logins, and tasks that failed to complete for many users.
This incident came as OpenAI was holding its annual DevDay developer conference in San Francisco, where the company was showcasing new products and capabilities for developers. OpenAI’s official status reporting later described the incident as affecting multiple services, including the Agents API.
This disruption adds another service-reliability test for OpenAI, as ChatGPT and its surrounding developer ecosystem become increasingly important to consumers, businesses, and developers alike.
ChatGPT users report service problems.
OpenAI initially acknowledged elevated errors affecting ChatGPT. The company said some users could experience failed requests, difficulty logging in or signing up, and tasks that didn’t complete properly.
Downdetector data cited by Forbes showed nearly 1,000 outage reports for OpenAI shortly before 2 p.m. EDT, with about 85% of those reports relating specifically to ChatGPT. The number of reports fluctuated sharply during the afternoon, suggesting the impact wasn’t uniform across all users.
That distinction matters here. An elevated-error incident doesn’t necessarily mean ChatGPT is completely unavailable to everyone. OpenAI’s status system reports availability at an aggregate level, while individual users can experience different results depending on their account, the model in use, and which service they’re accessing.
The incident extended beyond ChatGPT.
This outage was broader than the initial ChatGPT reports suggested. OpenAI’s incident description identified elevated errors across ChatGPT, Codex, and the API, with later updates specifically including the Agents API too. The reported symptoms included failed requests, login and signup difficulties, and tasks that failed to complete.
That wider impact matters, since OpenAI now operates far more than just a consumer chatbot. Developers use its APIs to build applications, Codex supports software-development workflows, and agent-related services increasingly connect models with external tools and business processes.
When several layers of that ecosystem experience elevated errors simultaneously, the disruption can affect users well beyond people simply trying to open ChatGPT.
The outage came during OpenAI DevDay.
The timing added another dimension to this incident entirely. OpenAI’s annual DevDay took place in San Francisco on September 29, bringing developers together for technical sessions, demonstrations, and discussions around the company’s latest tools.
The company was also using the event to showcase new AI capabilities. Reuters reported that OpenAI launched Dots, an always-on autonomous AI agent designed to pursue user goals across applications. This announcement positioned OpenAI more directly against other companies developing autonomous AI assistants and enterprise agents. Reuters also reported technical glitches during the live demonstrations themselves.
This simultaneous service disruption therefore occurred at a particularly visible moment for OpenAI. It also highlights a broader issue for AI providers: as AI systems become embedded in development tools, workplace software, and business applications, service availability becomes more important than simply the availability of a single chatbot.
Downdetector reports showed real volatility.
Third-party outage tracking showed that the scale of reported problems changed rapidly throughout the incident. Forbes reported that OpenAI-related reports on Downdetector approached 1,000 shortly before 2 p.m. EDT. Reports then fell below 100 before climbing back above 500 again roughly 15 minutes later.
Downdetector figures represent user-submitted reports rather than a direct measurement of OpenAI’s infrastructure. They’re therefore useful for showing the timing and visibility of a disruption, but readers shouldn’t treat them as an exact count of affected users.
OpenAI’s own status reporting provides the more important confirmation that service errors were actually occurring.
Claude also experienced an outage.
This disruption wasn’t limited to OpenAI, either. Anthropic’s Claude experienced a separate spike in outage reports earlier on September 29. Reuters reported that Claude-related reports peaked at more than 10,000 around 10:25 a.m. EDT, before falling sharply, while Anthropic said the issues had been mitigated by that point.
These simultaneous problems across major AI platforms illustrate how much demand and infrastructure now sit behind consumer-facing AI services generally. That said, there’s no evidence from the available reporting that the ChatGPT and Claude incidents shared a common technical cause. They should therefore be treated as separate, unrelated service events.
OpenAI’s reliability challenge is getting larger.
For OpenAI, reliability is becoming a broader business issue as its products keep expanding. ChatGPT has become the consumer-facing gateway to the company’s models, while APIs provide infrastructure for external applications. Codex and agent products add another layer by letting AI systems perform increasingly complex tasks.
That means an outage can have different consequences depending on which service gets affected. A failed ChatGPT request may interrupt an individual user’s work. An API disruption can affect an entire application built on OpenAI infrastructure. An issue involving agents can potentially interrupt automated workflows that depend on models completing multi-step tasks reliably.
The September 29 incident involved several of these layers at once, according to OpenAI’s status reporting.
OpenAI is also scaling its financial ambitions.
This outage came during a particularly important period for OpenAI’s business more broadly. On the same day, Bloomberg reported that OpenAI was seeking at least $30 billion in new funding at a potential valuation of around $1.4 trillion. These discussions were described as early and subject to change.
Reuters separately reported the same funding target and valuation, citing Bloomberg’s reporting. The proposed financing would serve as a bridge while OpenAI delays an anticipated public listing.
This funding report shouldn’t be interpreted as connected to the service outage. There’s no evidence in the available reporting that the two events are related in any way. Instead, they illustrate two different sides of OpenAI’s expansion. The company is pursuing increasingly large-scale infrastructure, products, and commercial opportunities, while its services are also becoming increasingly important to customers and developers at the same time.
What the outage means for businesses
For businesses building around AI, this incident reinforces the importance of treating model providers as genuine infrastructure dependencies. Companies using external AI APIs need to consider availability, fallback systems, monitoring, and contingency plans, just as they would with any other critical technology provider.
This issue becomes more important once AI agents are allowed to perform actions, rather than simply generate text. An agent that can’t reach its underlying model may stop a workflow, delay a customer interaction, or leave a multi-step task incomplete entirely.
OpenAI’s September 29 incident included the Agents API in its later status description, making that distinction particularly relevant for developers currently experimenting with agentic applications.
AI reliability is becoming part of the product itself.
This incident points to a larger change happening across the AI market. Early generative AI products were often treated as standalone tools. Today, AI is increasingly connected to software development, customer service, enterprise applications, payment workflows, research systems, and autonomous agents.
That changes what reliability actually means. For an ordinary chatbot, downtime is frustrating. For an AI service embedded inside a business process, downtime can become a genuine operational problem instead.
The September 29 outage didn’t permanently disrupt OpenAI’s services, and the available reporting doesn’t establish a major infrastructure failure or security breach behind the incident. What it does show is the growing operational dependence on AI services across multiple layers of the technology stack. As OpenAI expands from ChatGPT into APIs, coding tools, and autonomous agents, maintaining reliable access will remain an important part of that broader expansion.
What happened to ChatGPT, in summary
OpenAI reported elevated errors affecting ChatGPT on September 29, with users potentially experiencing failed requests, login or signup problems, and incomplete tasks. The incident was later described more broadly as affecting ChatGPT, Codex, and the API, including the Agents API.
This disruption occurred during OpenAI DevDay, when the company was simultaneously presenting new AI products and developer capabilities to the industry. The exact technical cause wasn’t established in the reporting reviewed for this article. OpenAI’s status reporting focused on identifying and mitigating the elevated errors, rather than publicly attributing the incident to a specific underlying cause.
For businesses and developers, the practical lesson is clear: as AI becomes infrastructure, availability matters just as much as model capability.
ChatGPT Down For Some Users As OpenAI Reports ‘Elevated Errors’


Comments
No comments yet. Be the first to share your thoughts.
Leave a comment