In September 2025, Nature published an editorial underscoring a striking gap: until now, no widely used Large Language Model (LLM) had undergone a formal peer-review process in a scientific journal. The launch of DeepSeek-R1, recently subjected to this rigorous evaluation, marks a precedent of great significance. For the scientific, medical, and academic communities, the debate is not limited to technical issues. It encompasses transparency, reproducibility, and legitimacy in a technology already permeating biomedical research, higher education, and clinical practice (Nature Editorial, 2025).
Peer review—long considered the cornerstone of modern science—provides methodological clarity and enables independent verification of claims. Without it, LLMs risk consolidating their role as “black boxes” adopted at scale without thorough examination of their limitations and biases.
LLMs and their impact on biomedical science
The relevance of Large Language Models (LLMs) extends far beyond computer science. Recent studies show their application in biomedical literature retrieval, clinical trial design, and even regulatory report drafting (Singhal et al., 2023). However, the reliability of these applications depends on transparent knowledge of training methods, datasets, and evaluation mechanisms. The absence of independent review undermines trust when these systems are applied in diagnostics, pharmacovigilance, or health policy.
The World Health Organization has consistently emphasized that emerging technologies must adhere to principles of transparency, equity, and accountability (WHO, 2021). Publishing an LLM through peer review aligns artificial intelligence with these principles, encouraging its ethical integration into sensitive domains such as hospitals, regulatory agencies, and academic institutions.
Between innovation and accountability
The DeepSeek-R1 case
DeepSeek-R1 stands out because its developers deliberately submitted the model’s architecture and results to independent peer reviewers. This decision not only legitimizes the technological advancement but also pressures the global community to demand equivalent standards from other major developers. The practice highlights a fundamental lesson: innovation without accountability risks undermining the credibility of the entire field.
Risks of opacity
LLMs can amplify biases embedded in training data, generate responses that mimic certainty without evidence, and propagate errors when integrated into medical or governmental systems. Without external review, these risks remain concealed until they result in real-world consequences: misdiagnoses, failures in critical medical translations, or inadequate policy recommendations.
Implications for medicine and scientific policy
In the pharmaceutical sector, an LLM subjected to peer review could become a reliable tool to accelerate molecule identification, analyze drug interactions, and systematize clinical trial data. For academia, it represents the possibility of integrating AI into teaching and scientific writing without compromising editorial quality standards. For governments, peer-reviewed evidence provides a stronger foundation for regulating AI, rather than relying solely on corporate communications.
Peer review does not guarantee perfection, but it does establish a transparent minimum standard of quality. Just as a scientific article gains legitimacy when published in Nature or The Lancet, a peer-reviewed Large Language Model gains credibility for use in biomedical research and hospital environments.
Toward a culture of open science in AI
Nature’s call to “bring your LLMs” into peer review suggests a broader challenge: embedding artificial intelligence within the culture of open science. This entails sharing not only results but also training datasets, evaluation metrics, and safety protocols. In doing so, the global community can replicate, challenge, and improve upon published findings.
Latin America and Europe play a crucial role in this shift. Latin America brings momentum in open-access policies, while Europe is building one of the most advanced regulatory frameworks for AI. Should both regions adopt peer-reviewed publication of LLMs as a standard, the ripple effect could shape the future of biomedical research worldwide.
Final reflections
The DeepSeek-R1 milestone opens a path that other developers of Large Language Models (LLMs) must follow. For professionals in medicine, research, and public policy, this development redefines the relationship between technological innovation and scientific validation. In a world where LLMs already draft clinical protocols, translate regulatory documents, and synthesize medical literature, demanding peer review is not an academic formality. It is a prerequisite for safety, ethics, and trust.
The opportunity is clear: to integrate artificial intelligence into science with the same rigor that governs biomedical discoveries. At Scienslate, we believe that clear communication and specialized translation are essential to ensure these technologies truly benefit global health and knowledge. Discover how our medical translation services can support your next international research project.
References
- Singhal, K., Azizi, S., Tu, T., et al. (2023). Large language models encode clinical knowledge. Nature, 620(7972), 172–180. https://doi.org/10.1038/s41586-023-06291-2
- World Health Organization. (2021). Ethics and governance of artificial intelligence for health: WHO guidance. https://www.who.int/publications/i/item/9789240029200
- Nature Editorial. (2025, September 17). Bring us your LLMs: why peer review is good for AI models. Nature, 645, 559. https://doi.org/10.1038/d41586-025-02979-9
