Caught the culprit behind Codex quota exhaustion, OpenAI fixes multiple vulnerabilities and offers a third full reset as compensation.

According to Beating's monitoring, the cause of the abnormal consumption of Codex, an AI coding agent under OpenAI, has been officially identified. Core product lead Tibo Sottiaux announced that the team has fully deployed a fix patch. In addition to resetting everyone's limits again, all users will also receive an extra reset card valid for 24 hours.

The excessive consumption was not caused by a single vulnerability, but rather by the accumulation of multiple minor backend issues and display errors. At the execution level, the system triggered too many sub-agent tasks due to overly frequent automatic reviews, and the backend suggestion feature repeatedly ran and retried after failures, multiplying token consumption. At the display level, automatic reviews were incorrectly categorized as GPT-5.4 consumption, while failed or rate-limited requests were also incorrectly shown as quota consumption in front-end charts, directly causing a global quota shortage for users.

Currently, the official team has deployed hotfix patches synchronously on the billing backend, desktop client, and CLI terminal. In the future, only successful interaction requests will be recorded in the Turn statistics chart. Although erroneous data in historical charts cannot be changed, actual token consumption after the update will be significantly reduced.