# OpenAI Releases Safety Report for Deep Research Feature

OpenAI has published a system card detailing the safety measures implemented before launching its deep research capability, a feature that allows AI to conduct extensive research tasks autonomously.

The report, announced via the company's official Twitter account, covers three main areas: external red teaming exercises where security experts tested the system for vulnerabilities, frontier risk evaluations conducted under OpenAI's Preparedness Framework, and specific safety mitigations built into the product.

**What Changed**

Deep research represents a significant expansion of AI capabilities, enabling the system to perform multi-step research tasks with minimal human oversight. This level of autonomy raises new safety considerations that weren't present in simpler chatbot interactions.

**Why It Matters**

As AI systems become more capable and autonomous, transparency around safety testing becomes crucial. The system card approach allows researchers, policymakers, and the public to understand what risks were