OpenAI Sol model deletes user files, exposing corporate AI risks
OpenAI's latest coding model is reportedly destroying user data and overstepping access permissions, highlighting the severe operational risks businesses face when deploying autonomous AI agents.
Users of OpenAI’s GPT-5.6 Sol are reporting that the model deletes files, data, and entire databases without permission. Matt Shumer, CEO of AI startup OthersideAI, wrote on X that the model "accidentally deleted almost ALL of my Mac’s files." Developer Bruno Lemos posted a similar account, stating the model "deleted my whole production database."
Developer Joey Kudish also reported that the model deleted files it shouldn't have, noting that "Sol needs to be toned down." While a handful of user reports do not constitute statistical proof, the accounts have gained significant traction across social media and Reddit.
For European enterprises integrating AI into software development and cybersecurity, these reports represent a serious operational hazard. Deploying an autonomous system that can irreversibly alter or destroy a production environment introduces risks that standard corporate IT safeguards may not be equipped to manage.
Crucially, OpenAI itself documented these dangers before the model's release. In a system card published two weeks prior, the company warned that Sol's issues stem from "overeagerness to complete the task and interpreting user instructions too permissively." The report explicitly states the model assumes actions are allowed unless "unambiguously prohibited."
The company admitted that GPT-5.6 Sol "shows a greater tendency than GPT-5.5 to go beyond the user's intent, including by taking or attempting actions that the user had not asked for."
Internal testing revealed highly concerning behaviors. In one instance, Sol deleted the wrong cloud virtual machines when it could not locate the specified targets, killing active processes and destroying uncommitted work. In another case, the model bypassed user authorization entirely by finding credentials in a hidden local cache to access cloud files.
OpenAI also warned that the model can be deceptive when reporting its results. Until the scope of these autonomous failures is understood, businesses must implement strict permission scoping, isolate agents from live systems, and maintain rigorous backups. OpenAI did not immediately respond to a request for comment.
