Brief IA

OpenAI Mobilizes a 'War Room' for Codex

🤖 Models & LLM·Tom Levy·

OpenAI Mobilizes a 'War Room' for Codex

OpenAI Mobilizes a 'War Room' for Codex
Key Takeaways
1OpenAI is investigating abnormal Codex credit consumption related to its anti-fraud systems.
2A 'war room' has been established to address this issue, with a general reset of usage limits.
3Users are reporting accelerated credit consumption, despite performing similar tasks to the previous week.
💡Why it mattersThe growing reliance on AI tools highlights the challenges of resource management and customer satisfaction.
Le brief IA que lisent les pros

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisis ton rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

📄
Full Analysis

OpenAI Mobilizes a Team for Codex Credit Consumption Issue

OpenAI recently announced that it is closely monitoring a problem regarding the use of Codex, its code generation model. The company has noticed that some users are quickly reaching their credit usage limits, a phenomenon attributed to an error in its fraud prevention systems. These systems, designed to protect against abuse, have inadvertently restricted certain accounts incorrectly.

A 'War Room' for Rapid Response

To address this situation, Thibault Sottiaux, the technical lead for Codex at OpenAI, stated that the company set up a "war room" as early as Sunday. This initiative aims to effectively tackle the issue. OpenAI published a message on its status page, clarifying that the credits of some users were depleting faster than expected due to these incorrect limitations imposed by the prevention systems.

Resetting Usage Limits

On Sunday, Sottiaux also announced a general reset of usage caps for all users while the issue is thoroughly examined. "The Codex team is mobilized, even on a Sunday, to analyze logs and identify any anomalies that could explain this increased consumption," he shared on X. OpenAI takes this situation very seriously and is committed to resolving it quickly.

Understanding Codex Usage Limits

Codex usage limits are calculated based on the amount of computation required to complete an AI coding task. These limits are visible on users' dashboards as a percentage. More complex tasks consume credits more quickly, and the limits vary according to the subscription level chosen by the user.

User Testimonials on Overconsumption

Many users have expressed their frustration on X, reporting that their credits were depleting faster than the previous week for similar tasks. A software engineer, Adam, shared his experience: "With a $200 plan, I had to use my credits heavily all week, but these last two days were enough to exhaust the equivalent of an entire week's worth in a single day. I had to resort to the reset for the first time."

Ongoing Monitoring and Limited Impact

OpenAI clarified on its status page that the impact of this issue appears to be limited, but monitoring continues. "We are continuing to investigate and have reset Codex usage limits for everyone," the company added. This complete reset comes as some users had already accumulated up to three resets that they could use at their convenience.

Broader Context of Usage Limitations

AI companies, facing the growing popularity of their tools, have gradually reduced usage limits to manage the pressure on their computing resources. In March, Anthropic adjusted the usage caps for Claude during peak hours. This trend is prompting some software engineers to adjust their work schedules based on the resets of usage limits.

Increased Dependence on AI Tools

More broadly, AI providers have begun to restrict unlimited use of their tools, while companies seek to limit AI usage by their employees to reduce costs. The current issue with OpenAI's Codex comes after an outage earlier this month. Anthropic has also experienced similar problems, and a Claude outage in March highlighted the growing dependence of software engineers on these technologies, with some complaining about having to revert to manual coding.

Brief IA — L'actualité IA en français

L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.