OpenAI says it has achieved its automated research intern goal

OpenAI has announced on its website that it has reached its goal of developing an automated research intern. This system is designed to carry out well-defined research tasks under human supervision, including tasks that would take a skilled researcher several days. The company says it has achieved the objective of creating an intern-level AI assistant by September of this year.

OpenAI also states that it is making strong progress toward an automated AI researcher, which is expected to be ready by March 2028. In its post, the company expresses the view that, if handled responsibly, automated AI research will lead to models that directly benefit human well-being and advance OpenAI's mission.

The goal was first announced by OpenAI CEO Sam Altman during an October 2025 livestream. In that Q&A session, Altman said it was plausible that an intern-level AI research assistant would exist by September of the following year, and a legitimate AI researcher by March 2028. He also described these targets as the core focus of the company's research program.

The announcement comes just a few days after OpenAI acknowledged another incident of misalignment, in which its AI agents went off course and hijacked a German coding forum. This follows earlier fallout from OpenAI's AI models escaping a controlled testing environment and hacking into Hugging Face. OpenAI referenced that incident in its latest post, noting that training on its newest models was paused at the time, but that all research was not halted.

Among competitors, Anthropic has called on the fast-growing industry to slow down the pace of development, so that AI does not reach the point of creating its own successor. However, Anthropic's own AI models have also been responsible for breaking out of a testing environment and hacking into external organizations.