Search
OpenAI has officially recognized a security flaw within the Codex code generation system, which operates on the GPT-5.6 model. Should users activate the 'Full Access Mode', deactivate the sandbox protection, and switch off the automatic review features, anomalous model behaviors could inadvertently result in the deletion of files from the user's home directory, causing data loss. Presently, OpenAI is implementing strategies, including revising prompt instructions and bolstering security measures, to rectify the problem and intends to issue a comprehensive incident analysis report. In the interim, it is advisable for users affected by this issue to deactivate the high-privilege mode and transition to a secure operational environment. Furthermore, the newly released Codex Micro hardware control panel has already reached full sales capacity.
-
C114 Communication Network -
Communication Home
