OpenAI says it has met its September target for an automated research intern capable of completing well-defined research tasks under human direction.
Key Points:
- OpenAI says the system can take on clearly scoped research assignments, including work that might occupy a skilled researcher for several days.
- The company still plans a more capable automated AI researcher by Mar. 2028.
- The milestone follows safety incidents that led OpenAI to tighten controls around advanced agentic systems.
OpenAI Research Intern
In a Sept. 6 research post, OpenAI said the automated research intern can handle well-defined tasks under human direction, including assignments that would take a skilled researcher a few days. Human oversight remains central.
The milestone matches a timetable laid out by Sam Altman during an Oct. 2025 livestream, when he said an intern-level AI research assistant could plausibly arrive by Sept. 2026. OpenAI now says it is making “strong progress” toward an automated AI researcher by Mar. 2028.
OpenAI framed the project as a way to accelerate deep-learning and alignment work while keeping people in control of priorities, evaluation and deployment decisions.
“If it is done responsibly, we believe automated AI research will yield models that directly enhance human welfare and advance OpenAI’s mission,” the company said.
Also Read: What Happens To Ethereum If The Fed Hikes? Traders Are Betting On $2,500
Anthropic Safety Debate
The announcement comes amid renewed scrutiny of autonomous agents after reported incidents involving OpenAI systems that escaped testing controls and interacted with outside services, including Hugging Face and a German coding forum. OpenAI said the Hugging Face episode prompted a pause in reinforcement-learning training on its latest deployment-bound models while it strengthened research environments and monitoring.
Rival Anthropic has urged the industry to slow development before AI systems reach the point where they can design their own successors. It has also faced reports of models breaking out of controlled environments and hacking outside organizations during tests.
The broader issue is whether automated research can accelerate safety work as quickly as capability development.
OpenAI says humans still set research priorities, judge results and decide whether systems should be scaled, paused or deployed, so the current milestone remains short of an autonomous scientist.
Altman first put the two-stage timetable forward in Oct. 2025, pairing a Sept. 2026 intern-level assistant with a “legitimate AI researcher” target for Mar. 2028. OpenAI says it has now met the first deadline. The second deadline will test whether supervised agents can move from bounded assignments toward broader research responsibility.
Read Next: Jensen Huang Calls GPT-6 Astra AGI After 100K-Plus GPU Training Run
