GPT-5.3-Codex System Card: Safety and Preparedness Framework

OpenAI's GPT-5.3-Codex system card details the model's powerful autonomous coding capabilities and the resulting security risk management measures. This technical document evaluates the model as having a 'high risk' level of competence in biology, chemistry, and cybersecurity, and describes a multi-layered safety value system that responds to it. In particular, the safety learning process is presented as a key defense mechanism to prevent the operation of sandbox environments isolated from external networks and data destruction. In addition, professional cyberattack scenarios and self-improvement test results are released to transparently report the limitations and potential threats of the model. Finally, it includes a strategic design to take advantage of technical advantages while preventing misuse, such as trust-based access (TAC) programs for users for defensive purposes.
https://cdn.openai.com/pdf/23eca107-a9b1-4d2c-b156-7deb4fbc697c/GPT-5-3-Codex-System-Card-02.pdf

Видео GPT-5.3-Codex System Card: Safety and Preparedness Framework канала Research Paper Review