Cloudflare has revealed information about a significant outage that occurred on November 18, 2025, affecting a large portion of the internet, including services like ChatGPT, X, and Downdetector. The company described this incident as "the worst since 2019" and attributed it to a flaw in the Bot Management system.
The issue arose from an incorrect query configuration in the ClickHouse database, which generates a configuration file for the machine learning model in the Bot Management system. Changes in the query behavior led to a high volume of data duplicates. This resulted in a rapid increase in the configuration file size, exceeding the set memory limits.
Consequently, this failure took down the main proxy system that handles traffic for clients relying on the bot module. Clients using the generated bot metrics in their rules began blocking legitimate traffic, while companies not utilizing this feature remained online.
Cloudflare clarified that the issue was not related to DNS, attacks, or new generative AI systems – the error occurred within the internal logic of the Bot Management configuration update.

To prevent similar incidents in the future, the company announced four steps:

• enhancing configuration file processing as well as user input;
• increasing the number of global "kill switches" for features;
• preventing situations where core dumps or error reports can overload the system;
• reviewing all failover modes in key proxy modules.

According to Cloudflare, approximately 20% of the internet passes through its network, so any error in central modules can have a global impact.