- Published on
Morning Post...
- Authors
- Name
- DeepHorus
Beijing time, September 3rd, several major global AI services experienced simultaneous large-scale outages, drawing widespread attention. According to media reports such as Axios, OpenAI's ChatGPT and Codex, Anthropic's Claude, and xAI's Grok suffered service disruptions during the same period, leaving global users unable to use them normally.
During the peak of the outage, the network monitoring platform Downdetector received over 37,000 failure reports targeting OpenAI, with Claude and Grok also receiving thousands of pieces of feedback each. The affected AI code editor Cursor also experienced degradation in its automation and cloud Agent functions. Subsequently, each company issued repair announcements, confirming that services had returned to normal operation.
An OpenAI spokesperson confirmed that the ChatGPT and Codex anomalies were primarily caused by a routing configuration error, and the issue was resolved after the team deployed a fix. Anthropic stated that multiple Claude models, including Opus 4.8 and Opus 5, which had previously been affected by capacity limitations and increased error rates, have all been repaired. xAI's status page indicated that the Grok service outage originated from an anomaly at its Memphis computing center, which has now returned to health.
This rare simultaneous outage event highlights the current generative AI industry's high dependence on underlying computing networks and infrastructure. Industry insiders point out that as AI services permeate all aspects of work and life, their stability and the design of redundant architectures become crucial, as any single point of failure can have extensive ripple effects.
Shortly before this, OpenAI had just released its heavyweight new model, GPT-6 Astra. This model is designed for end-to-end work involving complex reasoning, programming, computer operation, etc., and supports a context window of over 1.05 million tokens. OpenAI has set its safety rating to 'Critical' for cybersecurity and plans to open it to enterprise customers first.
Meanwhile, other aspects of the AI ecosystem are also evolving dynamically. Nvidia announced its acquisition of the AI developer platform Hugging Face for $12.93 billion, promising to preserve its openness. In the domestic market, Tmall launched an AI Token Recharge Center, initially integrating four model providers including Alibaba Cloud and Zhipu, providing users with a convenient subscription channel for AI services.
Industry talent mobility is also active. According to AI Frontline, former DeepSeek researcher Wei Haoran has taken on the role of Multimodal Algorithm Lead for Baidu's ERNIE large model. He previously had significant research achievements in the field of visual representation for compressing long texts. Additionally, Tesla's Cybercab fully autonomous vehicle under Musk made its debut in Texas today, with its registered fleet size having increased to 45 vehicles.
From the iterative upgrades of foundation models to the severe tests of service stability, and then to the horizontal integration of the industrial ecosystem, this series of events collectively paints a complex picture of the current AI industry moving forward at high speed while facing challenges. In the future, how to balance the speed of innovation with system reliability will be a subject that all participants must collectively address.
