OpenAI reset Codex usage limits across all tiers Tuesday. Thibault Sottiaux, the project lead known as Tibo, scrapped the caps after dashboards showed users hitting rate limits at an unexplained rate. Sottiaux called the decision “cautious” in a post on X. OpenAI simultaneously purged a group of fraudulent accounts to reclaim compute resources.
- OpenAI project lead Thibault Sottiaux resets Codex usage limits following an unexplained surge in users hitting rate caps on Tuesday.
- The company purged a group of fraudulent accounts to reclaim compute resources before a temporary 2x cap boost expires April 2.
- Sottiaux admits the engineering team lacks a full explanation for the usage spike as legitimate adoption competes with coordinated botnet attacks.
The dashboard surge appeared suddenly. Sottiaux admitted the team lacks a full explanation for the spike. “Our Codex dashboards are showing increased rate of users hitting rate limits and since we don’t fully understand why I have made the cautious decision of resetting the usage limits,” Sottiaux wrote.
The April 2 Catch
The relief is temporary. Internal logs and community reports point to a 2x allowance that expires April 2. Users noticed the change immediately. One social media post flagged a doubled cap appearing on their dashboard. Others confirmed limits were being drained “faster this week” before the reset occurred.
OpenAI is fighting two fronts. Legitimate adoption of AI coding tools is surging. At the same time, coordinated botnets are trying to steal compute. Estimates place Codex weekly active users in the millions. The company did not share details on the fraud methods detected or the total volume of recovered compute.
Capacity and Scaling History
OpenAI has a history of reactive limit tweaks. Resets happened in February following outages and the Codex Mac app launch. Those previous adjustments also included temporary boosts to headroom. The current April 2 deadline suggests this reset functions as a capacity buffer. OpenAI is likely investigating telemetry data before setting a permanent cap structure.
Genuine News Deserves Honest Attention.
High-conviction projects require an intelligent audience. Connect with readers who value sharp reporting.
👉 Submit Your PRDevelopers in production environments face a monitoring challenge. Production workflows relying on Codex could see performance shifts once the temporary boost expires Thursday. OpenAI has not commented on long-term capacity plans or hardware expansion to meet the current demand.
Compute as a High-Value Target
Recovering compute from a “pocket” of fraudulent accounts highlights a shift in cybercrime. AI compute has become a high-value commodity for hackers. These actors use stolen accounts to run large-scale inference or train smaller models on OpenAI’s infrastructure. Sottiaux framed the purge as a “shared victory.” However, the lack of detail on the “pocket” size leaves questions about the total scale of the abuse.
Usage patterns remain under heavy monitoring. The April 2 expiration date serves as a warning for the developer community. The extra headroom is a tool for investigation rather than a permanent upgrade to the service tiers.
Chain Street’s Take
This is damage control. Sottiaux admitted the team is flying blind on the hit spike. When an infrastructure provider says they “don’t fully understand” their own usage data, it suggests a massive telemetry failure.
Compute is the new gold. Purging a “pocket” of accounts is a band-aid on a systemic leak. The April 2 cutoff proves OpenAI isn’t ready to commit to this level of scale. The floor is shaky. Relying on Codex for high-uptime production work is a gamble while the team relies on these reactive resets to keep the lights on. OpenAI needs to fix its telemetry before it can claim a “shared victory” over fraud.
Activate Intelligence Layer
Institutional-grade structural analysis for this article.
