OpenAI GPT-5.5 System Card

This document is the GPT-5.5 System Card published by OpenAI, a technical safety and governance disclosure describing the capabilities, limitations, evaluation results, and risk mitigations associated with the GPT-5.5 large language model. The document states that GPT-5.5 was assessed against OpenAI's Preparedness Framework and that the model received 'medium' risk ratings across evaluated categories including cybersecurity, biological, chemical, nuclear, and radiological threats, persuasion, and model autonomy, with the document asserting that no category reached the 'high' threshold that would block deployment. The document discloses that GPT-5.5 exhibits measurable uplift potential in certain dual-use domains, including cybersecurity and persuasion tasks, and that mitigations including usage policies, system-level safeguards, and operator-level controls are described as the primary mechanisms for managing these risks, though the document does not provide granular technical specifications of those controls. The document engages with emerging AI governance frameworks, most notably the EU AI Act and voluntary commitments made under frontier AI safety frameworks, and the safety evaluation methodology described may be subject to evolving regulatory interpretation, particularly in the EU where the AI Act establishes mandatory conformity obligations for high-risk and general-purpose AI systems. Material compliance considerations include the adequacy of internal red-teaming as a substitute for third-party auditing, the sufficiency of disclosed mitigation measures under future mandatory transparency and safety-testing requirements, and the interaction of dual-use capability disclosures with export control and biosecurity regulatory regimes.