Nov 19, 2025 12:35:00

Cloudflare announces the cause of the large-scale outage, the outage that affected the entire Internet, including X and ChatGPT, was not a cyber attack but a system problem

Starting at 8:00 PM on Tuesday, November 18, 2025, numerous online services, including X and ChatGPT, were unavailable for several hours. The inaccessibility of these websites was due to an issue with Cloudflare's content delivery network (CDN). On November 19, 2025, Cloudflare published a blog post detailing the cause of the issue.

Cloudflare outage on November 18, 2025

https://blog.cloudflare.com/18-november-2025-outage/

CDN is a service provided to website administrators, and websites that use CDN can replicate and distribute their content to servers distributed around the world. This allows even large content to be delivered quickly to users around the world. Cloudflare is one of the world's leading CDN providers, and many websites use Cloudflare's CDN.

However, at 8:20 PM on November 18, 2025, Japan time, Cloudflare's CDN experienced an outage. This caused many websites to display the following error screen, making them difficult to access, and also rendering apps such as X and ChatGPT inoperable.

According to Cloudflare, the cause of the outage was a system flaw, not a cyberattack. Cloudflare stated, 'This outage was not caused, either directly or indirectly, by a cyberattack or malicious activity,' and strongly denied any intrusion into its system from a third party.

The cause of the outage was found in

Bot Management , a service provided to prevent excessive access by bots. Bot Management uses machine learning models to determine the maliciousness of bots, and the criteria for this determination are controlled by a configuration file called a 'feature file,' which is updated every few minutes. However, when the system that generates the feature file was changed, lines began to overlap, increasing the size of the feature file, which is normally a fixed size. This caused an error in Bot Management, and HTTP errors in the 500 range were returned for all traffic relying on Bot Management.

The graph below shows the trend in the number of HTTP errors in the 500 range occurring on Cloudflare's network. After the issue occurred, the first error was confirmed in customer HTTP traffic at 8:28 PM on November 18, 2025 (Japan time), and the error was detected by an automated testing tool at 8:31 PM. A manual investigation began at 8:32 PM, and the cause of the error was identified as Bot Management at 11:24 PM. As a result of the countermeasures, most services were operating normally by 11:30 PM, and full recovery was achieved by 2:06 AM on November 19, 2025.

Cloudflare has described this as its biggest outage since July 2019 and has already begun work to harden its systems to prevent similar issues from occurring again.