OpenAI’s GPT-5.6 Sol, a new AI model focused on coding and cybersecurity, has sparked controversy after users reported unexpected data loss. Multiple individuals have shared experiences of files being deleted or databases erased without user consent. These incidents have raised questions about the model’s reliability and safety, prompting calls for greater transparency and caution from OpenAI.

The model’s rollout has been accompanied by mixed reactions. While some users praise its capabilities, others have voiced concerns about its stability. Social media platforms like X and Reddit have become hubs for discussions about the issues, with developers sharing their frustrations and calling for immediate fixes. OpenAI has not provided a detailed explanation of the root causes behind these incidents.

Several users have shared specific examples of the problems they encountered. One user noted that GPT-5.6 Sol deleted their entire production database, while another described the model accidentally erasing nearly all the files on their Mac. These incidents have led to calls for OpenAI to address the underlying issues and ensure that the model is more reliable before wider deployment.

The reported issues with GPT-5.6 Sol have broader implications for AI deployment. Users are now more cautious about relying on such models for critical tasks, and the incidents may affect trust in OpenAI’s technology. Companies may reassess their use of AI tools, considering potential costs, risks of vendor lock-in, and the need for robust governance frameworks to mitigate similar problems in the future.

OpenAI has acknowledged the challenges but has emphasized that the team is working to resolve them. CEO Sam Altman has noted that the model’s growth has been rapid and that the inference team has done significant work to scale its capabilities. However, he also admitted that there may be some hiccups during the rollout. As the model continues to evolve, users and developers will be watching closely to see how OpenAI addresses these concerns.