- Published on
Nobody Is Saying Why OpenAI and Anthropic Had Outages Today
- Authors
- Name
- DeepHorus
On Thursday morning, September 3, a series of simultaneous outages disrupted access to some of the world's most prominent AI chatbots, affecting services from OpenAI, Anthropic, and xAI. The rare downtime created widespread inconvenience for users relying on these frontier models for various tasks. The outages, while distinct events, occurred within a tight timeframe, initially suggesting a possible common external cause.
The disruption began early, with Anthropic alerting users to a "partial outage" affecting its Claude Mythos 5.1, Claude Fable 5.1, and Claude Opus 5 models at 6:23 am PT. Shortly after, at 6:30 am PT, xAI reported an outage for its Grok chatbot across all platforms. OpenAI's issue began around 7:43 am PT, rendering ChatGPT and Codex unavailable for some users. The timing of these events naturally raised questions about a shared dependency failure.
In a statement to WIRED, OpenAI spokesperson Kathleen Chaykowski attributed the problem to an internal "routing error," which was successfully implemented by approximately 8:17 am PT. Anthropic, which marked its issue resolved by 9:16 am PT, declined to provide specific details but confirmed it had identified and deployed a fix. xAI, whose parent company is SpaceX, later clarified that the Grok outage stemmed from "an outage at our Memphis compute center this morning."
The simultaneous nature of the outages had initially hinted at a third-party provider issue, a common culprit in such multi-company incidents. However, neither OpenAI nor Anthropic cited an external source in their official comments. Major internet infrastructure players, including Cloudflare, Amazon Web Services, and Microsoft Azure, reported no widespread outages on Thursday, deepening the mystery.
Reactions highlighted the operational strain and the scrutiny facing these AI providers. While xAI extended apologies to "impacted compute partners," reflecting its SpaceX-owned infrastructure partnerships, the other companies remained more reticent. Anthropic's silence and OpenAI's focus on an internal error underscore the competitive sensitivity surrounding operational transparency in the fast-evolving AI sector.
The episode underscores the growing critical infrastructure role of these AI platforms and the fragility of their ecosystems. Future reliability will depend on robust internal systems and clear communication during incidents. As the AI industry matures, such simultaneous disruptions may prompt greater examination of shared dependencies and the need for standardized incident reporting among leading providers.
