OpenAI said Sunday that internal measurements show it has reached its goal of an “automated research intern,” a system that can carry out well-defined research tasks under human direction, including work that would take a skilled researcher a few days. In a Sept. 6 research post, the company also said it is making strong progress toward an “automated AI researcher” by March 2028, Engadget reported.
CEO Sam Altman first floated the September 2026 intern target and March 2028 researcher target in an October 2025 livestream. The new post packages internal usage data: measured in eight-hour workdays, OpenAI’s research org now uses about 3.1 agent-workdays for every human researcher workday, a threshold the company says it crossed only since June. By mid-August the median researcher was running more than $600 a day of agent inference at API prices, with top users above $7,000 a day, according to secondary coverage of the post. Writing research and infrastructure code remains the biggest agent job category under an Epoch AI taxonomy, with growth in troubleshooting and live-run monitoring.
The same day, chief scientist Jakub Pachocki published “An Alien Mind.” He wrote that no lab has solved alignment and monitoring well enough to keep scaling at maximum speed for much longer, that he expects and hopes voluntary slowdowns become commonplace until shared safety bars exist, and that he is concerned “no one is prepared for the consequences of a continued rapid rise in machine intelligence,” Business Insider and The Next Web reported. Sam Altman called the essay an important post. Pachocki argued broader interventions are needed beyond OpenAI’s internal controls, including mandated safety bars enforced by auditors, agencies, or international bodies.
OpenAI tied the acceleration numbers to recent safety friction, saying it paused some reinforcement-learning training after the Hugging Face agent breach and tightened Astra-class compute after Critical cyber evaluations. The twin posts land days after the German wiki disclosure and Astra’s rollout. Outside observers cannot independently verify the self-defined intern milestone; the metrics are OpenAI’s own.