DeepSeek hit by global outage as mystery Harness account surfaces on Bilibili
DeepSeek's chat service and two API tiers went down worldwide on the morning of October 15, with the company saying it had identified the cause and was repairing it. The roughly hour-long disruption coincided with the quiet appearance of a "DeepSeekHarness" account on Bilibili, though no official evidence links the two.
The outage began at about 8:11 a.m. Beijing time, according to the incident record published on DeepSeek's status page. The company first said its team was investigating the problem and identifying the cause, then updated the incident status to "identified" and moved into the repair stage. By the time the latest update was posted, the disruption had lasted about an hour.
The services listed as affected are DeepSeek's chat service, the DeepSeek V4 Pro API and the DeepSeek V4.1 Flash API. The status page carried a red failure notice stating that DeepSeek's web pages and API were partially interrupted.
Users described different symptoms. Some sent messages and received no reply, some were stuck on a loading screen, and others ran into prompts reading "server busy" or "too many messages, please try again later." The overseas service-monitoring account Status Is DOWN quickly flagged the anomaly and asked whether other users were affected.
The third-party monitoring platform StatusGator showed 86 related reports over the previous 24 hours, a figure that was still rising. A user in Luxembourg said the server was not responding, a user in Poland reported "no reply," and a user in New York was more concise: "Too slow."
Previous DeepSeek outages have set off a familiar routine in comment sections: first checking whether one's own network is at fault, then confirming that DeepSeek is indeed down, and finally speculating that a new model is on the way. During a brief anomaly a few days earlier, an overseas user asked on Reddit whether DeepSeek was "on holiday again." Some commenters there said they could not use the service either, some worried aloud about losing their chat history, and some connected the server trouble directly to an expected DeepSeek-V5.
This time the timing drew attention for another reason. While DeepSeek's servers were failing, an account named "DeepSeekHarness" appeared on Bilibili. Its public page shows a newly created account that has posted little information so far, but it already has more than 10,000 followers.
DeepSeek Harness is an open-source agent runtime platform the company has already released. In effect, it fits a model with an "exoskeleton" for doing work: the model handles the thinking, while Harness calls tools, edits files, executes commands and processes code, and can even split complex tasks among multiple sub-agents. DeepSeek's stated core positioning for it is "Everything is a plugin." The Bilibili account therefore looks more like preparation to promote the agent product to users in China.
Whether the sudden service failure was preparation for a Harness update or a new model release, there is no official evidence either way. The two events did land at the same time: web pages and APIs stopped working together, and the Harness homepage on Bilibili went up quietly. The new model has not appeared; the stage, as the report put it, has been set first.