OpenAI says it hit its own "automated research intern" target -- with its chief scientist warning about what comes next
The research org now logs 3.1 "agent-workdays" per human workday, but more than half of successful multi-hour agent tasks still needed a human to step in -- and the same announcement carries a warning from chief scientist Jakub Pachocki that AI risk is "going to grow from here."
OpenAI said on September 6, 2026 that it has reached a goal CEO Sam Altman first floated in an October 2025 livestream: fielding an "automated research intern" by September 2026, with a further goal of a "legitimate AI researcher" by March 2028. In a blog post on what it calls research acceleration, the company said its research organization is now logging 3.1 "agent-workdays" of effort for every workday put in by a human researcher, describing the system, per Engadget's Jackson Chen, as "a system that can carry out well-defined research tasks under human direction."
What "intern-level" actually means here
OpenAI is explicit that this isn't an autonomous scientist setting its own research agenda -- it's a supervised agentic system executing a bounded objective a human hands it, then returning the work for evaluation. The company's own numbers show the limits of that supervision: on tasks estimated to take 4 to 8 hours, more than half of successful attempts over the past six months still involved at least one human intervention along the way, per Fortune's reporting on the same post. That's a real capability jump from a year ago, and also a long way from research running itself.
The spend behind the milestone, and the warning attached to it
The infrastructure cost of getting here is substantial and uneven across OpenAI's own research staff: by mid-August 2026, the median researcher was running more than $600 a day of inference at API prices, with the heaviest users past $7,000 a day, according to Fortune's Jeremy Kahn. The same announcement carried a less triumphant note from chief scientist Jakub Pachocki, who warned that "risks associated with AI are unfortunately going to grow from here" as models get better at working around the chain-of-thought monitoring safety researchers currently rely on to check their reasoning -- and said he expects voluntary slowdowns to become common until the industry has shared safety standards.
Grading your own homework
Every capability figure in this announcement -- the 3.1x ratio, the intervention rate, the inference spend -- comes from OpenAI measuring OpenAI's own internal research process against a target it set for itself a year earlier, with no independent benchmark or outside auditor in the loop. That doesn't make the claim false, but it does mean the milestone should be read as a lab's own progress report rather than third-party verification that the "automated research intern" bar has actually been cleared. That the same post pairs the achievement with its own chief scientist's warning about growing risk is worth taking at face value too -- it's not the framing a company reaches for if the goal were purely to oversell the result.