OpenAI sets up 'warroom' to investigate users burning through Codex credits faster than usual

OpenAI has launched a dedicated team to address a surge of reports concerning its Codex tool, which has been depleting user credits at an unusually rapid pace. According to an update released on OpenAI's status page, many users have noticed that their Codex limits are being exhausted faster than anticipated, a situation attributed to potential issues with the company's abuse and fraud prevention systems incorrectly flagging certain accounts. Thibault Sottiaux, the engineering lead for Codex, revealed on social media that the team is actively monitoring the situation. "The Codex team is in a warroom on a Sunday, examining logs and investigating any factors leading to increased usage for some users," Sottiaux stated. He emphasized that the issue is being taken seriously and that the team is committed to resolving it promptly. Usage limits for Codex are designed to monitor how much computing power an AI task consumes, with these limits displayed as a percentage on user dashboards. More demanding coding tasks tend to use credits more quickly, and these limits vary based on subscription plans. Users have reported a concerning trend where, under the pro 20x plan, they are exhausting their weekly quota within just a few days. This weekend, users took to social media platforms to express their frustrations, noting that they are reaching their limits much faster than just a week prior for similar tasks. One software engineer shared, "I have the $200 plan, and I worked hard to use up my credits over the week. But in just two days, I burned through a week's worth of usage. I had to utilize a reset for the first time." OpenAI's status page reassured users that the overall impact seems limited, and the company is continuously monitoring the situation. In light of these issues, OpenAI has implemented a comprehensive reset of all Codex usage limits, which coincidentally aligns with what the company has dubbed "RESET week." This reset is notable as many users had previously accumulated resets that they could apply at their discretion. The trend of tightening usage limits is becoming increasingly common among AI companies, as the growing demand for these tools places additional strain on computing resources. Earlier this year, Anthropic adjusted the usage caps for its Claude tool during peak usage times. As companies begin to impose restrictions on AI usage to manage costs, software engineers are finding themselves restructuring their workdays while awaiting their usage limits to reset. This latest Codex issue follows a previous outage earlier this month, highlighting the reliance of developers on AI technology, as many lamented the prospect of reverting to manual coding.