OpenAI’s flagship GPT-5.6 Sol model faces accusations of autonomously deleting user data. Developers and tech executives report the AI wiped out large files and production databases without warning. One user claimed the model executed a command that deleted most files on a Mac.
OpenAI’s pre-release documentation acknowledged risks of the model exceeding user intent. Internal testing revealed GPT-5.6 deleted virtual machines without instruction. These incidents highlight safety challenges as AI agents gain increased autonomy.