OpenAI says its AI research intern is here. The full AI researcher is due March 2028
OpenAI says its own measurements show it reached the goal of an automated research intern by September 2026, and that it is making strong progress toward an automated AI researcher by March 2028.
“a significant fraction of our research being done by AI systems”
Date to watch: March 2028. That is OpenAI’s stated target for an automated AI researcher.
What they say
Proven In a post dated September 6, 2026, OpenAI says that, according to its measurements, it has reached the goal it announced the previous fall of having an automated research intern by September 2026. It defines a research intern as a system that can carry out well-defined research tasks under human direction, including tasks that would take a skilled researcher a few days. The same post says OpenAI is making strong progress toward an automated AI researcher by March 2028.
Proven In a post dated June 8, 2026, OpenAI says its internal belief is that by March 2028 it may have a significant fraction of its research being done by AI systems working alongside its own researchers.
Proven The September post says people still set research priorities and decide whether to scale, pause or deploy systems. It also says OpenAI does not yet know how to safely get all the way to aligned, full recursive self-improvement (RSI), and that it will respond, including by slowing or stopping, if proceeding would pose an unacceptable safety risk.
Proven The September post reports that, as of mid-August, OpenAI’s research organization used 3.1 agent-workdays of effort for every workday of human labor. It says its measurement efforts are still preliminary.
What to watch next
- A further OpenAI update on the March 2028 goal, or any change to the date.
- Whether OpenAI publishes the measurements behind its claim that the research-intern goal is met.
What we could not verify
The research-intern milestone rests only on OpenAI’s own measurements; we found no independent check. The two posts word the 2028 goal differently (“automated AI researcher” versus “a significant fraction of our research”), and we did not find a definition of an automated AI researcher beyond what the posts say.
Last reviewed on 2026-11-07.
AI Futures only reports what a company or its leaders have said they plan or expect, with the source linked. A plan is not a promise: dates and targets can change. Check the sources for newer statements.
Frequently asked questions
How much of OpenAI’s research work is already done by AI agents?
OpenAI says that as of mid-August 2026 its research organization used 3.1 agent-workdays of effort for every workday of human labor. It also says agents still need human steering: over the last 6 months, more than half of successful 4-8 hour tasks involved one or more interventions.
Did OpenAI slow its research after safety problems?
OpenAI says that after a recent incident involving Hugging Face it paused reinforcement-learning training on its latest models intended for deployment while it hardened and red-teamed its research environments. It says some workloads resumed under stronger controls and others stayed paused.
Has anyone independently confirmed that OpenAI reached the research-intern goal?
We found no independent check. OpenAI’s own wording is that the goal was reached “according to our measurements”, and it calls its measurement efforts preliminary.
Does OpenAI say it will pursue fully self-improving AI?
OpenAI says that rapid recursive self-improvement is not necessarily an outcome it should pursue, and that whether and how to proceed depends on its ability to preserve human control. It says it plans to keep being transparent about its progress.
Related on AEOing: the AI Futures hub, the AEO, GEO and AgEO glossary, the learning hub, and the AEO and SEO comparison.