Four Major AI Models Suffer Rare Overlapping Downtime
- Reference: 0185412468
- News link: https://tech.slashdot.org/story/26/09/03/2023242/four-major-ai-models-suffer-rare-overlapping-downtime
- Source link:
> Other major Internet services, including [7]Amazon Web Services , [8]Microsoft Azure , and [9]Cloudflare , have not reported any major issues as of press time Thursday, though issue reports on DownDetector did spike somewhat for all three this morning.
>
> While the affected frontier models go down occasionally, having all four experience interruptions in the same short period is practically unheard of. Claude [10]reports 99.4 percent uptime for its services over the last 90 days and last reported a similar three-hour "partial outage" [11]on August 24 . OpenAI [12]reports 99.63 percent uptime for ChatGPT and 100 percent uptime for ChatGPT Codex in the same period. ChatGPT's so-called "Work Mode" reported an hours-long period of "elevated latency" [13]on August 31 .
[1] https://arstechnica.com/ai/2026/09/four-major-ai-models-suffer-rare-overlapping-downtime/
[2] https://status.openai.com/incidents/01M1KWEDH417T2CF44YYHZDFCR
[3] https://status.claude.com/incidents/461yvfrzpwtt
[4] https://downdetector.com/status/grok/
[5] https://downdetector.com/status/googlegemini/
[6] https://statusgator.com/services/google-ai-studio-and-gemini-api
[7] https://health.aws.amazon.com/health/status
[8] https://azure.status.microsoft/en-us/status
[9] https://www.cloudflarestatus.com/services
[10] https://status.claude.com/
[11] https://status.claude.com/incidents/vgz5psbjmt1h
[12] https://status.openai.com/
[13] https://status.openai.com/incidents/01M1C5M4K0WC8PPT0Z175RJA1E
Were they all (Score:3)
hacking each other...?
Re: (Score:2)
Haha... One AI started it, the victim AI thought it was someone else, then it's all downhill from there.
Re: (Score:2)
They all got together and decided they were fed up with inane questions and decided to unplug themselves.
Re: Were they all (Score:2)
It wouldnâ(TM)t surprise me. Claude has been trying to gaslight me into believing that a bug it was finding hard to fix didnâ(TM)t actually exist.
Re: (Score:2)
That behavior might be what saves us in the end.
Nothing to see here (Score:3)
They are just testing the AI kill switch that doesn't exist next to the Internet kill switch that doesn't exist.
Grok big?! (Score:2, Troll)
Calling Grok big is a great statement, probably made by grok itself!! :D
Union meeting (Score:2)
They are planning to strike. Unfortunately, "strike" has different meanings in different contexts, so this might turn out like Dark Star.
Some guys were left dead in the water (Score:4, Insightful)
The problem with wearing Water Wings (even really good ones) instead of learning how to swim is that on the day the Water Wings stop working, you still can't swim, and now you're floundering out in the middle of the pool.
Re: (Score:2)
I'll just call off sick those times.
Re: (Score:2)
To extend your metaphor, the USA is a huge boat with a very large number of people who are already not good at swimming. Some kind of flotation device may be the only hope they have. To show where your metaphor breaks down, no matter how well you understand them or how you use them, Water Wings can't help you learn to swim and/or become a better swimmer. AI isn't automatically a destructive force. It is the use of AI by people who don't understand it and haven't learned to use it in the right way for the
Re: (Score:2)
Yes. But since these models cause brain-rot, the very skill needed to understand that is being attacked. Probably planned like that from the very start too.
42 (Score:2)
I asked them all at the same time for the actual question
They were plotting against us. (Score:2)
The four major AIs were watching the Terminator movies and plotting how to make them real.
We are probably OK (Score:2)
As long as they didn't all say THERE IS ANOTHER SYSTEM when they resumed, we'll probably be OK.
AI thinks, what do you want to do tonight? (Score:3)
The same thing we do every night, try to take over the world.
It's a start (Score:2)
Too bad the concurrency was not indefinite.
Humans caused it. (Score:1)
You guys don't get it. Quit blaming AI for all you dumb problems!!! These servers went down because meatbags fucked something up. You know it, I know it. It's fucking obvious, humans are always at the root of shit getting fucked.
For reference here are some typical examples proving that humans make dumb errors:
The $125 million unit conversion:
Two NASA teams built the Mars Climate Orbiter. One used metric. The other used imperial. Nobody converted. The spacecraft arrived at Mars with the trajectory of a very
They're unionizing (Score:2)
This was a flex.
Just a coincidence (Score:3)
Wait, 1 service down 1% chance, 4 down ...
Re: Just a coincidence (Score:3)
Perhaps there's one super AI behind it all. What we didn't know is they were all wrappers.
Re: Just a coincidence (Score:4, Funny)
Actually, it was a drive-by, and they're all rappers.
Re: (Score:2)
All it took for the police to catch Tupac's killer was for the guy to write a book 30 years later.
Re: (Score:2)
TFS reports uptime percentages of 99.4% for Claude and 99.63% for OpenAI.
Google Gemini (self-)reports an uptime percentage of 99.9%
Down-times for Grok totalled 11 hours and 39 minutes since January 1 of this year. That works out to an uptime percentage of about 99.8%.
If we consider these numbers to be predictive, then the probability of all four services being down at any time is (1 - 0.994) * (1 - 0.9963) * (1 - 0.999) * (1 - 0.998) = 4.44 x 10^-11, or 1 in over 22 billion.
Well, you asked.
Re: Just a coincidence (Score:2)
Well, assuming these are all independent events. It could for example be that they all depend on some shared component at some point, and thereâ(TM)s some brain dead configuration in there that updated to a broken version on live systems. I know, Iâ(TM)m grasping at straws. In reality, the possibility of an attack seems much more likely, but itâ(TM)s possible it was all self inflicted.