Beneficial
Deployments

More than half of the world’s population—an estimated 4.5 billion people—still lacks access to essential health services, according to the WHO. In low- and middle-income countries, poor-quality care is estimated to contribute to roughly 5 million deaths each year, while another 3.6 million people die because they never reach care at all. Sub-Saharan Africa carries over a quarter of the world’s disease burden but has just 3% of the global health workforce. We seek to address these disparities at three levels: alongside clinicians at the point of care, with health ministries across whole populations, and with the wider field through shared, open infrastructure.

At the primary care level, we are working to improve patient outcomes in the US and around the world by deploying AI to help address some of the biggest challenges in the system. We believe AI can help ensure patients can receive high-quality care, for example by supporting clinical decision-making and integrating with the latest point-of-care diagnostics. We work closely with Ministries of Health and clinicians to ensure that any care AI supports is safe, effective, and held to clinical standards of evidence.

At the population level, Claude can help health ministries and public health officials bring together fragmented health and population data to make faster and more informed decisions, including around workforce deployment, supply chain management, and disease surveillance and response.

And for the wider field, we are creating a suite of digital public goods that any ministry, partner, or model can adopt. By releasing these openly, we hope to set a high standard for AI in healthcare and extend these efforts beyond what we could deploy alone. As part of this work, we have joined global health and technology actors to support the Open Health Stack (OHS) Software Foundation in creating open-source building blocks for digital health solutions.

Within Anthropic’s broader life sciences efforts, our beneficial deployments focus on the diseases and patients the market tends to overlook. Funding for drug development is largely informed by potential market size, which leaves behind patients suffering from less common diseases and those unable to access or afford treatment. An estimated 300 to 400 million people live with one of more than 7,000 rare diseases, yet fewer than 5% of those diseases have an approved therapy, and most patients wait five to seven years for a diagnosis. Hundreds of millions more contract neglected tropical diseases each year, yet disease surveillance is too thin to guide intervention, and existing therapies are often directed at diseases that circulate in wealthier countries. By collaborating with clinical researchers, patient organizations, and data scientists through grants, hackathons, and other research initiatives, we aim to gain a better understanding of where AI can address these challenges.

It’s already possible to develop diagnostics or drugs for many neglected diseases because the underlying mechanisms are well understood. Unfortunately, this is not true for many other illnesses, such as Alzheimer’s disease, mental health conditions, and various kinds of cancer. Here, we must continue to do the kind of basic research that broadens the scientific frontier. Modern biological research generates massive amounts of data, yet turning that data into validated findings still depends on manual processes that can’t easily scale alongside the rate at which data is produced. Through our partnerships, we’re exploring the use of AI agents to collect and analyze data, and throughout the process of scientific discovery.

To push the frontier of science, we also are making broad investments in model training and product development, including Claude Science, an AI workbench for scientific research. Over the past months, our research teams have made substantial progress on deepening specific scientific capabilities in domains such as chemistry, bioinformatics, protein design, and more.

And given that transformative discoveries emerge through experimentation, we’re putting Claude in the hands of as many scientists as possible through our AI for Science grant program, which offers usage credits to qualified researchers, and a Claude Team plan for Scientists, which offers principal investigators discounted access to apps like Claude Science.

In past waves of technological innovation, the positive impacts were concentrated in a few regions and occupations before spreading slowly and unevenly. In contrast, AI is changing how we work at an accelerated pace. Anthropic’s Economic Index tracks AI’s impact on the labor market, while our Economic Futures program studies potential policy responses. We build alongside partner organizations to help millions of workers and businesses succeed in the AI transition, while building the evidence base for future policy interventions at scale.

Our first focus area is putting AI to work for workers. Families lose out when the systems meant to support them are too complex to navigate. In the US, for example, an estimated $140 billion in government benefits go unclaimed every year. We’re partnering with organizations such as NextLadder Ventures, Code for America, and the National Domestic Workers Alliance to develop Claude-powered solutions that help workers navigate these systems and improve their economic security. We are also working with states like Maryland to improve benefits access and strengthen the safety net to better support workers in this AI transition.

Second is supporting workers as they build the skills and agency to thrive in the AI economy. With CodePath and Claude Corps, for example, we are developing new models for how workers can rapidly learn and apply AI skills as jobs evolve.

Third is enabling builders and their businesses. In the US, more than 36 million solo entrepreneurs and small and mid-sized businesses account for nearly half of America’s private sector employment, yet these businesses face persistent barriers to capital and constant administrative friction. With solutions like Claude for Small Business, we aim to help business owners manage their operations. And through partnerships with institutions like Pacific Community Ventures, we’re piloting projects that use AI to help businesses access capital investment and support services faster.

Today, more than half of children in low- and middle-income countries can’t read a basic text by age 10. Socioeconomic achievement gaps have, in many cases, widened over the past 50 years, with the US seeing some of the largest in the OECD. We believe AI can help close these gaps and prepare students for a life of flourishing and agency by bolstering the work of educators, schools, universities, and the organizations that support them.

Our work in K-12 education begins with supporting teachers. We aim to make evidence-based practices available to every classroom, and reduce the planning and administrative load that keeps teachers from using them. Products like Claude for Teachers are designed to give educators more time to focus on the creative parts of their jobs and connecting with students. And we’re providing Claude licenses, training, and community support to Teach For All’s global AI Literacy and Creator Collective, which helps teachers across 63 countries build AI fluency and create tools that are relevant to their communities.

We’re also exploring how AI can help extend high-quality instruction. One-on-one human tutoring is among the most reliably effective interventions in education research, but it is also among the most expensive to deliver. With our partners, we are researching ways to integrate AI into tutoring that reduce its cost but retain its efficacy. We share educators’ concerns about cognitive offloading and skill erosion, so we prioritize research on when and how AI tutoring can be most helpful, model training and product improvements to resist such cognitive offloading, and deployments that keep teachers involved in the learning process.

At the university level, educators are considering how AI might speed up scientific research, and how they can prepare students for a fast-changing labor market. We partner closely with institutions like Western Governors University to study these areas, including new approaches to learning and credentialing that keep up with the pace of change. In parallel, Claude for Higher Education brings Claude to campuses with a tutor-like “learning mode” for students and training resources for educators.