How AI Giants Are Building Real Laboratories
Jakarta, 05 November 2026 – Not long ago, AI in life sciences meant a tool that helped researchers write up results faster or scan the literature for relevant papers. Useful, but clearly in a supporting role. That is no longer the full picture.
In 2026, two of the world’s leading AI companies, OpenAI and Anthropic, made moves that crossed a fundamental line: they stopped building software assistants for scientists and started building scientific infrastructure itself, including physical labs, robotic systems, and acquired biotech companies. The message is clear: AI vendors are no longer content to sit alongside the scientific process. They want to be inside it.
In April 2026, OpenAI released GPT-Rosalind, the first in a dedicated life sciences model series. The name is a deliberate nod to Rosalind Franklin, the crystallographer whose X-ray diffraction work was essential to decoding DNA’s double-helix structure and whose contributions were historically underrecognized. The choice signals intent: this is a model meant to do serious science, not generate science-flavored text.
GPT-Rosalind is built for chemistry, protein engineering, and genomics, three domains where the ability to reason about molecular behavior, sequence relationships, and experimental design matters far more than general-purpose fluency. OpenAI has reported benchmark performance on BixBench and LABBench2, evaluations designed specifically to test AI on real laboratory reasoning tasks. These are not the kind of benchmarks you see on a general AI leaderboard; they are designed to expose whether a model understands scientific problems or simply pattern-matches on surface features. Access to GPT-Rosalind is restricted to qualified customers through a trusted-access program, a deliberate choice that positions it as scientific infrastructure, not a consumer product. The architecture of that access program tells you something important: OpenAI is treating this the way you would treat a regulated professional tool, not a general-purpose chatbot.
In September 2026, Anthropic confirmed what had been the subject of considerable speculation in scientific circles: the company operates a physical wet lab in the San Francisco Bay Area. This is not a computational biology team running simulations on servers. It is a laboratory where biological experiments are physically conducted, a space that handles real samples, real reagents, and real equipment. For a company whose identity has been built around AI safety research, this is a meaningful expansion of scope.
The lab’s stated purpose is to explore how Claude can direct robotic laboratory equipment, not by suggesting what a human should do next, but by actively supervising the machines themselves: adjusting parameters, interpreting results in real time, and deciding what the next experimental step should be. That is a materially different relationship between AI and science than anything we have seen before at scale. What makes this disclosure particularly credible is what Anthropic chose to say alongside the announcement. The company’s head of life sciences described AI automation of lab execution as early-stage. A spokesperson explicitly clarified that the lab is not specifically designed for drug discovery. In a sector where announcement culture tends toward the grandiose, that level of candor about current limitations is notable and, from a scientific standpoint, more reassuring than inflated claims would be.
Beyond building internally, both companies moved to acquire existing biotech expertise.
Anthropic acquired Coefficient Bio in April 2026 for $400 million. Coefficient Bio’s work sits at the intersection of machine learning and biology, precisely the capability gap Anthropic needed to make its wet lab ambitions credible. The acquisition brought in scientific talent and proprietary methods that would take years to develop from scratch.
OpenAI acquired Orchestra Bio in October 2026. Orchestra Bio builds AI agents that coordinate drug-development operations. It does not just analyze data; it actively manages the sequence of tasks, decisions, and handoffs that make up a drug development pipeline. OpenAI is folding Orchestra Bio’s technology directly into the Rosalind platform, which suggests a clear strategic intent: Rosalind is not just a scientific reasoning model, it is meant to become an operational system capable of running drug development workflows end-to-end.
Together, these four moves (GPT-Rosalind, the Anthropic wet lab, Coefficient Bio, and Orchestra Bio) point toward what researchers call closed-loop integration: a system where AI reasoning, scientific tools, robotic execution, and experimental feedback are all connected in a continuous cycle, with no human bottleneck between insight and experiment.
Step back from the individual announcements and a coherent picture emerges. If closed-loop AI systems (where a model reasons about a scientific question, directs a robot to run the experiment, interprets the result, and immediately designs the next test) can be made reliable; the implications for life sciences are genuinely transformative. Protein engineering, genomics, drug development, and food science all share a common bottleneck: the time and cost of physical experimentation. A hypothesis that takes three months and a team of researchers to test in a traditional lab could, in principle, be tested in days with AI-directed automation. That is not incremental improvement. That changes what is scientifically possible.
That said, scientific credibility requires distinguishing between what has been announced and what has been validated. As of November 2026, most publicly available information about these programs consists of press releases and corporate announcements, not peer-reviewed publications, independently verified throughput data, or disclosed experimental results. The gap between a confirmed wet lab and a proven high-fidelity automated research system is not trivial. Lab scale, operational capacity, and acquisition terms are all largely undisclosed. Ambition and execution are different, and technology history includes many well-resourced projects that took far longer than anticipated to deliver what they promised.
The honest summary: the race is real, the capital is real, and the direction is clear. But the moment when AI-directed laboratories routinely produce peer-reviewed, independently validated scientific breakthroughs at scale has not yet arrived. The professionals best positioned in this landscape will be those who can evaluate these claims with scientific rigor, those who understand enough about both biology and AI to know what a result actually means and whether the evidence behind an announcement justifies the conclusion being drawn from it.
At i3L University, we prepare students for careers at the intersection of science, technology, and industry. The developments described in this article, GPT-Rosalind, Anthropic’s wet lab, the Coefficient Bio, and Orchestra Bio acquisitions, are not distant industry news. They are actively reshaping the fields our graduates will enter in the next few years.
If you are studying in the sciences, the question is no longer whether AI will play a role in laboratory research. It already does. The more important question is, how do these systems actually reason about biological problems? Where are their genuine capabilities, and where do they still fall short? How do you design an experiment, interpret a result, or evaluate a published claim when an AI system helped produce it? These are becoming core scientific competencies, as foundational in their own way, as learning to use a pipette or read a gel.
If you are studying business and management, the Coefficient Bio and Orchestra Bio acquisitions are case studies in real time. They illustrate how capital flows between AI and industries, how competitive moats are being built through M&A rather than organic R&D alone, and how the traditional pharmaceutical and life sciences industry value chain is being disrupted from outside. The companies that understand this convergence early will shape it. Those that don’t will navigate it reactively.
The laboratory of the future is not a concept in a speculative report. It is being built right now, with real money, real scientists, and real equipment. i3L graduates, equipped with rigorous training and the ability to think critically across both science and business, have every reason to be among the people shaping what it becomes.
i3L University, formerly known as Indonesia International Institute for Life Sciences, offers interdisciplinary life sciences and business programs with global faculty, research, and innovation.
i3L University offers seven undergraduate programs: Biomedicine, Biotechnology, Pharmacy, Food Technology, Food Science & Nutrition, International Business Management, and Innovation & Entrepreneurship, and one master’s program of Master’s in Biomanagement.
