OpenAI Expands Deep Research Access & Releases Safety System Card

OpenAI has expanded access to its Deep Research feature, making it available to ChatGPT Plus, Team, Edu, and Enterprise users while introducing key improvements like embedded image citations and enhanced file handling.

This rollout comes alongside the release of a new Deep Research system card, which outlines the safety measures, risk assessments, and governance reviews conducted before deployment. Together, these updates reinforce OpenAI’s commitment to both innovation and responsible AI development.

What is Deep Research?

Deep Research is an agentic AI capability powered by an early version of OpenAI’s o3 model, optimized for web browsing. It is designed to:

• Search, interpret, and analyze vast amounts of online text, images, and PDFs.

• Adapt dynamically, pivoting based on newly encountered information.

• Read user-provided files and execute Python code for data analysis.

By utilizing these capabilities, OpenAI envisions Deep Research being useful across a wide range of applications, from academic research to investigative journalism.

Rigorous Safety Testing Before Release

Before rolling out Deep Research to ChatGPT Pro users, OpenAI conducted extensive safety evaluations, including:

• Preparedness Evaluations: Testing potential risks associated with AI autonomy.

• Governance Reviews: Ensuring compliance with OpenAI’s safety principles.

• Enhanced Web-Browsing Safeguards: Strengthening protections around personal information found online and training the model to resist harmful instructions encountered during searches.

Additionally, OpenAI used human probing and automated testing to refine its risk assessment methods, improving overall system security.

Deep Research System Card: Key Takeaways

The newly released Deep Research System Card provides a comprehensive overview of OpenAI’s testing, risk identification, and mitigation strategies before deployment. It highlights:

• Key risk areas, including privacy, bias, prompt injections, and hallucinations.

• Preparedness Scorecard assessments, ranking risks in areas such as cybersecurity, persuasion, and model autonomy.

• Mitigation strategies that address incremental risks posed by Deep Research’s web-browsing capabilities.

The system card rates risks like cybersecurity, persuasion, and model autonomy as 'Medium,' meaning the model meets OpenAI’s safety standards for deployment but still requires ongoing monitoring and refinement.

Deep Research Rolls Out to More Users With Key Improvements

OpenAI has officially expanded access to Deep Research, making it available to ChatGPT Plus, Team, Edu, and Enterprise users. This rollout brings enhanced AI-powered research capabilities to a wider audience, allowing more users to leverage Deep Research for complex, multi-step information gathering and analysis.

Since its initial launch, OpenAI has introduced several key improvements to Deep Research, including:

• Embedded images with citations to provide clearer, more reliable sources in responses.

• Better handling of uploaded files, allowing for more accurate understanding and referencing of user-provided documents.

Usage Limits by Subscription Tier

To ensure balanced access, OpenAI has set monthly Deep Research query limits based on user tier:

• Plus, Team, Enterprise, and Edu users: 10 Deep Research queries per month.

• Pro users: 120 Deep Research queries per month.

These enhancements reinforce OpenAI’s commitment to advancing AI-assisted research while expanding availability to more users across different plans.

What This Means

The expansion of Deep Research marks a significant step in OpenAI’s strategy to enhance AI-assisted research tools. By making it available to ChatGPT Plus, Team, Edu, and Enterprise users, OpenAI is broadening access to AI-driven multi-step research capabilities.

At the same time, the Deep Research system card highlights OpenAI’s focus on safety and governance, detailing the testing, risk assessments, and mitigations implemented before launch. These updates reinforce OpenAI’s commitment to both accessibility and responsible AI development.

With improvements like embedded image citations and better file referencing, Deep Research is becoming a more reliable tool for researchers, analysts, and professionals. As AI-powered research continues to evolve, OpenAI’s approach will help shape industry standards for secure, transparent, and effective AI-driven information gathering.

Editor’s Note: This article was created by Alicia Shapiro, CMO of AiNews.com, with writing, image, and idea-generation support from ChatGPT, an AI assistant. However, the final perspective and editorial choices are solely Alicia Shapiro’s. Special thanks to ChatGPT for assistance with research and editorial support in crafting this article.