Grok: A Bug Makes Responses Unintelligible for Some
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
Users of Grok Lite have been reporting unreadable responses generated by the xAI chatbot since Wednesday morning. The official account has confirmed a "rare and temporary" generation bug and advises users to restart the conversation, while stating that services are "operational." The phenomenon appears to be confined to Grok.com, and it has not been possible to reproduce the issue. In the background, xAI has lost a significant portion of its teams and launched a new model in July.
xAI has lost key profiles and claims to have launched a more efficient model
According to The Information, xAI has experienced a significant turnover in its workforce in recent months, losing the majority of its founding team as well as at least 50 researchers and engineers. In July, the company announced that it was launching a new base model, described as belonging to the "Opus class," offering faster execution, reduced token consumption, and lower costs.
Incoherent responses reported, especially on Grok Lite, since Wednesday
A bug has caused Grok to produce incomprehensible texts for many users. One user, after requesting a PDF generation, received several paragraphs of incoherent words, while another found links pointing to reinforcement learning research sites. Those who reported the issue say they are using Grok Lite and claim to have observed it since Wednesday morning. The Reddit community has been flooded with complaints; refreshing the session often restores the service, but some report that the anomaly persists despite several attempts.
Bug confirmed by Grok, limited scope to Grok.com and workaround advised
The malfunction seems to be limited to requests made on Grok.com, with the service's account on X not being affected. The Grok account confirmed the issues on X, describing them as a rare temporary generation bug, and apologized. It recommends starting a new chat or regenerating the response, indicating that this generally resolves the incident, while directing users to a status page where services are reported as fully operational. The bug could not be reproduced, and it is possible that its scope is limited to a portion of users. xAI has not responded to requests for comment.
Brief IA — L'actualité IA en français
L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.