INTRODUCING GPT-OSS-SAFEGUARD

Summary from the TLDR AI newsletter

The gpt-oss-safeguard models (120b and 20b parameters) let developers apply custom safety policies at inference time rather than training their own classifiers on thousands of labeled examples, using chain-of-thought reasoning to explain decisions.