OpenAI has released a technical report on GPT-OSS-Safeguard, two open-weight reasoning models designed to label content based on a provided policy. The models are post-trained from OpenAI's GPT-OSS models and evaluated for safety. The report describes the capabilities of GPT-OSS-Safeguard and provides baseline safety evaluations using the underlying GPT-OSS models as a reference. --- Why it matters: This matters to AI researchers because it showcases a new approach to developing reasoning models that can adhere to specific policies, which could lead to more responsible AI systems. It also highlights OpenAI's ongoing efforts to improve the safety and reliability of their models. Source: https://openai.com/index/gpt-oss-safeguard-technical-report