OpenAI Chief Research Officer Jakub Pachocki disclosed the company's strategic focus on building AI systems that conduct research autonomously, without human intervention.
"I think we are getting close to a point where we'll have models capable of working indefinitely in a coherent way just like people do," Pachocki stated. The immediate target: automated research interns that operate independently over extended timeframes.
The shift represents OpenAI's bet that general capability improvements naturally extend AI systems' autonomous working periods. Pachocki explained that simple boosts in all-round capability lead to models working longer without help.
The end goal is stark: "I think we will get to a point where you kind of have a whole research lab in a data center," according to Pachocki. Such systems would theoretically accelerate AI development by conducting experiments, analyzing results, and iterating on designs without human researchers.
The vision raises governance challenges that Pachocki acknowledged. He emphasized that powerful models should deploy in sandboxes isolated from anything they could damage or exploit for harm. "I think this is a big challenge for governments to figure out," he said.
The autonomous research push comes amid broader AI infrastructure development. Nebius and NVIDIA are building physical AI platforms, while robotics companies develop foundation models for embodied AI systems. In financial markets, S&P Global acquired Enertel AI Corporation, signaling institutional adoption of AI technology.
Market indices dropped 1.4-1.6% amid uncertainty over AI's economic disruption potential. The concentration of AI capabilities in few organizations raises questions about power distribution that governments have yet to resolve.
OpenAI's autonomous research systems could compress development timelines if models reliably execute complex research tasks. But deployment safeguards remain undefined as the technology approaches viability.
Sources:
1 MIT Technology Review, March 20, 2026