OpenAI Says It Has Reached Its Automated Research Milestone

OpenAI says it has reached an important milestone in its effort to automate artificial intelligence research, with its researchers increasingly using AI agents to handle complex research and coding tasks. The company says these systems are contributing to faster experimentation and are becoming more successful at completing difficult assignments.

According to OpenAI, the current stage resembles an “automated research intern,” with researchers using coding agents throughout their daily work. These agents can work on multiple tasks simultaneously, contribute code and assist with experiments, although human researchers continue to decide which ideas are worth pursuing and when systems should be scaled or deployed.

The company says its longer-term goal is considerably more ambitious: developing a fully autonomous AI researcher. Current reporting indicates that OpenAI believes it could reach that level in roughly 18 months, potentially around March 2028. The timeline is an internal target rather than a guaranteed prediction, and the company acknowledges that AI research contains many bottlenecks that cannot necessarily be accelerated at the same rate as individual tasks.

OpenAI also argues that automated research could have benefits beyond simply making AI development faster. The company believes increasingly capable systems could assist with alignment research, AI safety and the development of defensive technologies. At the same time, OpenAI has acknowledged that faster automated research could create significant risks if human oversight and control do not keep pace with improvements in AI capability.

The development represents a major shift in how AI companies approach AI research itself. Instead of AI systems simply helping researchers with individual tasks, increasingly capable agents could eventually participate in much larger portions of the research process. OpenAI says decisions about whether and how to proceed with increasingly autonomous systems must depend on maintaining meaningful human control and carefully weighing the benefits and risks.