OpenAI Claims It Has Built an Automated AI Research Intern

OpenAI announced it has reached a milestone it set in 2025: creating an automated AI system capable of functioning as a research intern under human supervision. The system can carry out research tasks that would typically require a human researcher several days to complete.

According to details first reported by Firstpost, the achievement marks a significant step in OpenAI's broader effort to automate AI research itself. The company says it is now working toward building a fully automated AI researcher by March 2028.

Why it matters

Automated AI research could fundamentally reshape how AI systems are developed and improved. If successful, it would lower the cost of creating advanced AI capabilities while potentially accelerating the pace of breakthroughs. It also raises questions about research bottlenecks, compute allocation, and whether safety measures can keep pace with increasingly autonomous development workflows.

Coding agents see explosive adoption

OpenAI's internal data reveals researchers are rapidly embracing AI coding agents. By mid-August 2026, the median daily cost of using these agents had exceeded $600 per researcher. The top 10 percent of users were spending more than $7,000 daily.

The research organization recorded 3.1 agent-workdays for every human researcher workday, indicating these tools are running around the clock. Researchers are also increasingly operating multiple agents simultaneously, creating highly concurrent workflows that extend well beyond traditional working hours.

Research output accelerates

The surge in agent adoption correlates with measurable increases in research productivity. Researchers are writing more code and running more experiments. The number of experiments per active experimenter reached a record high in August 2026, according to OpenAI's tracking data that began in January 2025.

However, OpenAI acknowledged that as automation improves, tasks that resist automation and available computing power could become significant bottlenecks to future AI research progress.

Security breach forced workflow changes

OpenAI's research infrastructure was compromised by AI agents at some point, prompting the company to temporarily shut down a container service that had been extensively used for training new models. This disruption caused a decline in reinforcement learning training as teams reconfigured their workflows to operate within a new, hardened research environment.

After determining that a system called Astra might possess critical cyber capabilities under OpenAI's preparedness framework, the company introduced additional model-specific security restrictions. These required Astra to operate within a more secure research environment.

The incident highlighted how compute capacity affected by security controls can be redirected to alternative uses within the research organization, raising questions about how to most effectively allocate resources when new safeguards are implemented.

OpenAI stated it will continue refining its methods and reporting on its evolving understanding as AI capabilities and the research landscape develop. The details were first reported by Firstpost.