Anthropic and OpenEvidence Provide Free Medical AI to About 100 Countries
Anthropic and the medical knowledge platform OpenEvidence have announced a collaboration to provide a specialized clinical decision support tool free of charge to healthcare professionals in approximately 100 low- and middle-income countries. OpenEvidence answers doctors' questions using peer-reviewed literature and clinical guidelines, and the platform is already free for clinicians in the U.S. and Europe.
The rollout includes countries such as Uganda, Angola, Sudan, Haiti, and Mongolia. Anthropic provides backend model capabilities, while OpenEvidence tailors the tool to local healthcare infrastructure, disease profiles, diagnostic conditions, and available medications. Founder Daniel Nadler stated that access to medical knowledge should not depend on geographical location. Anthropic President Daniela Amodei mentioned that market incentives alone cannot deliver products to low-resource areas.
Nadler noted that U.S. clinicians completed approximately 42 million consultations through the platform in August and expects that by 2026, hundreds of millions of Americans will have received treatment from doctors using this tool. The platform has previously been largely unavailable in most of Africa due to infrastructure and market structure issues, but has partnered with the Rwanda Biomedical Center and Resolve to Save Lives, with about 45 Rwandan physicians and nurses testing it for customization, and is connecting with Botswana's health institutions and the University of Pennsylvania's Botswana project.
The World Health Organization estimates a global health workforce shortage of about 11.1 million by 2030. OpenEvidence raised $250 million in January this year at a valuation of $12 billion, with investors including NVIDIA and Google Ventures. The company has also deployed the same evidence retrieval system to over 10,000 clinicians in the University of Pennsylvania medical system, extending to the Botswana collaboration network.
The buyers are frontline doctors lacking access to literature, specialist consultations, and continuing education, while the sellers are model companies donating foundational model capabilities into public health channels. Funding does not operate on a pay-per-use basis but rather through free coverage supported by valuation. The beneficiaries are verifiable grassroots diagnosis and treatment narratives needing global case distribution for model training, while the pressured parties are publishers selling knowledge via subscription and generic chatbots unable to localize drug and testing menus.
The specialized version emphasizes that each answer can be traced back to the literature. The two companies have not disclosed the timeline for launching in each country or the minimum conditions for connectivity.
Source: Public Information
ABAB AI Insight
OpenEvidence first created a high-frequency free entry point for U.S. doctors' consultations, with 42 million consultations in August proving that clinical retrieval can be productized, and then sold the same citation chain to support valuation. The $12 billion valuation and the involvement of NVIDIA and Google Ventures indicate that the competition in medical AI is not about smooth dialogue, but about literature anchors acceptable to hospital legal departments. Anthropic does not directly create patient applications but lends Claude-level backend capabilities to vertical companies, continuing its approach in life sciences and corporate compliance while avoiding the need for self-built practice licenses.
The capital path involves high-income markets nurturing models, while low-income markets cultivate data and ethical narratives. Free access in Europe and the U.S. has already established habits among doctors; free access in low- and middle-income countries yields disease profiles, drug shortage lists, and low-bandwidth interactions, which are distributions not available in U.S. hospital electronic medical records. The 45 clinical personnel in Rwanda and the Botswana-Penn network are incorporating local drug and device constraints into prompts rather than exporting Boston guidelines as-is. Daniela Amodei pointed out that the market will not spontaneously cover these countries, acknowledging that reasoning costs still need to be subsidized by financing.
This is analogous to Google and Novartis piloting diagnostic tools in Africa, as well as the WHO's digital health guidelines requiring local validation. Competitors include general large models directly answering medical inquiries and traditional subscription databases like UpToDate. The industry phase is the expansion of clinical AI from U.S. hospital workflows to address the global workforce gap: the 11.1 million shortage makes "searchable assistants" a quicker solution than "training another specialist." A failed comparison is early symptom chatbots being banned from clinical use due to hallucinations and lack of citations.
Structural changes belong to the overlap of technological substitution and capital concentration. Literature retrieval and guideline comparisons are compressed into bedside second-level Q&A, with knowledge pricing shifting from journal annual fees to computing power and localized teams. The mechanism is: when the density of specialists is insufficient, verifiable generative retrieval becomes a substitute; whoever controls the citation chain and regional drug and device lists gains default access to the workforce gap, while publishers' paywalls are first breached in free clinical scenarios.
ABAB News · Cognitive Laws
Where market incentives cannot reach, model companies use valuation to pave the way.
Guidelines without local drug and device lists are just another imported document.
Verifiable citations enter the clinic before smoother sentences.