# OpenAI Releases Safety Report for New o1 Reasoning Models

OpenAI has published a system card detailing the safety measures implemented before launching its o1 and o1-mini models, the company's latest AI systems designed for advanced reasoning tasks.

The report, announced via the company's official Twitter account, describes comprehensive safety evaluations conducted prior to release. These included external red teaming exercises—where outside experts attempt to find vulnerabilities—and frontier risk assessments aligned with OpenAI's Preparedness Framework.

**What's New**

The o1 models represent OpenAI's push into more sophisticated reasoning capabilities. Unlike previous models, these systems are designed to "think" through problems more deliberately before responding. The system card provides transparency into how OpenAI evaluated potential risks specific to these enhanced capabilities.

**Why It Matters**

As AI models become more powerful, safety documentation becomes increasingly critical. The system card offers researchers