SAN FRANCISCO—Anthropic is pushing into the healthcare market as it launches artificial intelligence tools and resources purpose-built for providers and payers.
The foundation AI model company’s Claude for Healthcare offering comes on the heels of Claude for Life Sciences, which it launched in October. The announcement, made in conjunction with the kickoff of the J.P. Morgan Healthcare Conference, marks Anthropic’s continued foray into regulated, high-stakes clinical environments.
Claude for Healthcare features HIPAA-ready infrastructure for enterprise customers, models trained specifically for healthcare and life sciences tasks and native integrations to commonly-used medical and scientific databases, including the CMS Coverage Database, ICD-10 codes and PubMed, according to the company.
The AI safety and research company also announced new capabilities for life sciences, extending from preclinical R&D into clinical operations and regulatory affairs.
The new capabilities mark Anthropic’s first foray in healthcare, but the company clearly sees big opportunities in the space.
Healthcare and life sciences represent one of Anthropic’s largest bets, Eric Kauderer-Abrams, head of biology and life sciences at Anthropic, told Fierce Healthcare.
The company's move into healthcare follows OpenAI's launch of ChatGPT Health last week, which is a hub to let users upload their medical records, as well as OpenAI for Healthcare as a suite of tools for healthcare enterprises.
Anthropic, founded by former OpenAI executives and senior research staff in 2021, developed a family of large language models called Claude.
Claude for Healthcare is built around Anthropic’s existing AI models but with connections to tools like medical databases and coding systems that make medical information easier to find, access and understand. With a HIPAA-ready infrastructure, healthcare organizations can deploy Claude for workflows involving protected health information like prior authorization for the first time, according to executives.
“These are not new models or even new products. What we're announcing is a coherent set of connectors so that Claude our models and Claude our products, however you're accessing Claude, whether it's through the chat interface or through Claude code or our enterprise offering, have the connections to the tools that professionals in healthcare and life sciences are using every day so that it can accelerate all the work they're doing, have all the context that they need to have and be present as a collaborator throughout the whole process, and function as a real amplifier of human capabilities,” Kauderer-Abrams said.
He added, “Our approach is to build useful tools and connections and make sure that our models are at a superhuman-level in all these use cases for payers, providers and health systems and for consumers as well. For life sciences, it’s about making sure that our models have skills spanning everything from the early-stage discovery through translation and commercialization, and that they connect to the tools that scientists and life science professionals are using every day.”
Kauderer-Abrams stressed that Anthropic is laser-focused on supporting AI for enterprise customers in healthcare rather than consumer health applications.
The company asserts that its commitment to building safe, responsible AI makes it uniquely suited to work with health systems, hospitals, payers and other healthcare organizations.
“Anthropic is a very natural fit for the healthcare and life sciences world because our identity as an AI company is built around safety and responsibility and rigor and reproducibility, and these are all the central tenants of the healthcare and life sciences industries,” he noted.
There are a number of pharma companies, health systems and startups already using Claude’s suite of LLMs—Banner Health, Stanford Healthcare, Novo Nordisk, Sanofi, AbbVie and Genmab—to automate administrative workflows like clinical documentation, regulatory submissions and clinical trial analysis.
“We were drawn to Anthropic's focus on AI safety and Claude's Constitutional AI approach to creating more helpful, harmless, and honest AI systems,” said Mike Reagin, chief technology officer at Banner Health, in a statement.
“Safety is non-negotiable in healthcare. Anthropic has been a clear leader in building models with strong safety foundations,” said Justin Norden, M.D., co-founder of startup Qualified Health.
“A big piece of the challenge of using AI in healthcare is there's basically no room for error. The cost of hallucinations, for example, or even the cost of non-reproducible analyzes is very high. We’ve done a lot of work to solve those problems,” Kauderer-Abrams said.
Anthropic uses specific technical methods during production model training to reduce the frequency of hallucinations and works to ensure that the results pulled from medical databases are accurate, cited and reproducible, he noted.
Anthropic boosted its life sciences AI leadership when it brought Kauderer-Abrams on board six months ago. The executive’s resume includes co-founding several computational biology startups.
With Claude for Healthcare, enterprise healthcare integrations now enable Claude to pull information from industry-standard systems and databases, meaning that clinicians and administrators can save time finding the data and generating the reports they need, executives said. Claude now connects to the Centers for Medicare & Medicaid Services (CMS) Coverage Database, including both Local and National Coverage Determinations. This enables Claude to verify coverage requirements, support prior auth workflows, reduce claim denials and surface regional coverage differences, according to the company.
Claude also connects to the International Classification of Diseases, 10th Revision (ICD-10) to look up both diagnosis and procedure codes to support medical coding, billing accuracy and claims management; the National Provider Identifier Registry, to help with provider verification, credentialing, networking directory management and claims validation; as well as PubMed, which provides access to more than 35 million pieces of biomedical literature.
The company also added new Agent Skills for FHIR Development to help developers connect healthcare systems faster and with fewer errors, improving the interoperability of medical data, executives said.
Claude for Healthcare’s HIPAA-ready infrastructure gives healthcare organizations a “plug-and-play” solution, Kauderer-Abrams said.
“A big part about being useful in healthcare and life sciences is respecting the intense compliance requirements that are there, and with good reason. We’re offering our solutions in a HIPAA-compliant way so that you can build applications around them, and so that you can do so in a way that respects all the relevant regulations. We're also the only major foundation model that's available through all of the major cloud providers,” he said.
The new AI tools available through Claude can support healthcare startups building new products as well as large enterprise health systems looking to integrate AI deeply into their operations, the company claims. By quickly pulling relevant information from coverage requirements, patient records and clinical guidelines, these AI resources can help organizations with time-intensive, burdensome administrative tasks like prior authorization requests and claims appeals and can also support care coordination and triage patient messages, Anthropic executives said.
“It’s really about amplifying the capabilities of the professionals in these healthcare organizations, so allowing you to be more productive, and allowing AI to take some of the grunt work off your plate and do it more efficiently and quickly. There is a huge amount of headroom to be had in terms of lessening the administrative burden,” Kauderer-Abrams said. “It's about the model being useful end-to-end, throughout all the processes, connecting to the tools that you use in your enterprise, operating in it natively as you do.”
Claude for Healthcare also includes consumer health integrations to enable individuals to share their health information as well as fitness and wellness data from different apps. Individuals can use Claude to summarize their medical history, understand test results in plain language, spot patterns across fitness and health metrics or prep questions for doctor appointments, according to Anthropic.
In the U.S., Claude Pro and Max plan subscribers can give Claude secure access to their lab results and health records. New HealthEx and Function connectors are now available in beta, while Apple Health and Android Health Connect integrations are rolling out in beta this week via the Claude iOS and Android apps.
Providers and healthcare companies have long had ambitions to collect patients’ health information all in one place—a longitudinal health record that makes it easier for patients to manage their health. But it’s been a bumpy road. Microsoft and Google both launched personal health record offerings but then shut them down.
Kauderer-Abrams is bullish that AI can finally make this happen.
“I’m excited about this idea of taking all of the context that you have about patients that are scattered throughout all these different places and having one system that can consolidate it all and connect natively into all the tools that are present in these healthcare systems to make care delivery more effective and more efficient,” he said. “Maybe we were all a bit misguided in trying to put all of the health at records in one place. That's too hard. Make the models go and do the hard work of figuring out how to interface and communicate and make sure all the connections are in place. That’s why I'm optimistic that the problem is solvable now is that there's been this unlocking intelligence.”
On the life sciences side, Anthropic is adding connectors to clinical trial solutions company Medidata, ClinicalTrials.gov, life sciences preprint servers bioRxiv & medRxiv, Open Targets, ChEMBL and Owkin.
The existing life sciences connectors include Benchling, 10x Genomics, PubMed, BioRender, Synapse.org and Wiley Scholar Gateway.
“On the life sciences side, I'm really excited to get to the point where you can have Claude as that collaborator through all stages of the R&D process and where Claude can take on increasingly large chunks of work and be increasingly autonomous,” Kauderer-Abrams said.