What happened
On September 3, several leading US AI services experienced simultaneous outages or elevated error rates:
- OpenAI ChatGPT: roughly 15 ChatGPT/Codex components reported errors, detected around 10:58 AM ET, with major degradation until about 2:56 PM ET before restoration
- Anthropic Claude: acknowledged partial outages, restored within hours and identified a cause
- xAI Grok: also affected, services returned to normal within hours
- Google Gemini: DownDetector showed a clear rise in interruption reports
During the outage, logins, Work Mode, file uploads, image generation, APIs and developer flows were temporarily disabled.
Root cause unknown
No provider has confirmed a shared root cause. Speculation ties the failures to OpenAI's upcoming Astra model or to cloud vendors such as Cloudflare or Microsoft Azure — none proven.
Takeaway
For organizations that depend heavily on a single AI provider, this was a real and costly operational-risk reminder:
- Keep multi-model redundancy and fallback plans for critical workloads
- Shared infrastructure means failures at major players can propagate to each other
Source: Bloomberg / Particle News / 9to5Mac / Ars Technica
No comments yet.