Calling major artificial intelligence companies labs may create a misleading impression of what these organizations have become. The Atlantic argues that companies such as OpenAI and Anthropic now operate more like powerful technology corporations than traditional scientific laboratories, despite continuing to embrace the lab label.
The terminology has historical roots. Industrial research institutions such as Bell Labs produced major technological advances, while MIT’s Laboratory for Computer Science helped establish foundations for modern cryptography. Early OpenAI and Anthropic similarly concentrated heavily on research. But both have since commercialized rapidly, reaching enormous valuations and generating billions of dollars in revenue.
The lab description can also provide a reputational advantage. Scientists generally enjoy greater public trust than business leaders, while technology companies face criticism over privacy, child safety, job displacement, data centers, and their growing economic influence. Associating AI companies with scientific research can therefore emphasize their stated public-interest missions while drawing attention away from commercial incentives.
Sun questions whether their research practices justify the terminology. After examining 100 recent research items listed by OpenAI and Anthropic, he found that most had no corresponding peer-reviewed paper accepted by an academic conference or journal. Although lengthy review processes and legal concerns can discourage technology companies from following conventional academic practices, independent scrutiny remains central to scientific credibility.
The article acknowledges that AI companies sometimes investigate problems that could reflect poorly on their own products. Anthropic, for example, has researched undesirable model behaviors. Still, company-funded research can raise questions about independence when organizations are studying technologies they also develop and sell.
Sun concludes that AI companies wanting to retain the lab designation should follow stronger scientific practices, including independent peer review, preregistered studies, and open-source code. More accurate terminology, he argues, could help the public recognize these organizations as corporations with enormous influence over workplaces, education, and society.
