During the recent gathering titled “The Briefing: AI for Science,” Anthropic unveiled Claude Science, an innovative “AI workbench for scientists” designed to consolidate scattered resources and data collections while producing charts and graphics. The firm, already leading with its coding aids and advanced models, positioned the release around AI’s capacity to “dramatically accelerate the pace of scientific discovery and the development of healthcare interventions,” and highlighted numerous biotech and pharmaceutical partners already relying on Claude.

Anthropic took an additional step by declaring plans to create medications internally. Life sciences lead Eric Kauderer-Abrams noted the focus would center on treatments for “neglected” conditions.

AI firms have pursued science and pharma clients aggressively, with OpenAI, Amazon, Google, and peers offering comparable life sciences platforms. Yet Anthropic’s initiative stands out as one of the bolder public efforts by a leading AI developer to pursue drug creation directly. This places the company in the distinctive spot of offering software to other drug developers that may compete with it. Anthropic enters a crowded field featuring AI-native firms such as Insilico, Google DeepMind’s Isomorphic Labs offshoot, emerging biotechs, and major pharmaceutical players developing or acquiring AI capabilities.

The company shared minimal concrete information on its drug development ambitions. At the event, Kauderer-Abrams avoided outlining next steps if promising candidates emerge. Anthropic declined to address follow-up questions from The Verge about initial disease targets or potential collaborations for laboratory work, animal studies, human trials, or production.

AI is applied at “every single stage of drug discovery.”

Specialists noted that the vagueness in Anthropic’s strategy mirrors wider questions about the AI drug wave overall. “AI drug discovery” covers a wide spectrum. It “is a really broad term,” said Namshik Han, a University of Cambridge professor and cofounder of AI biotech startup CardiaTec. AI supports “every single stage of drug discovery,” he noted, spanning compound identification and refinement through research support, data review, trials, and production. All leading drug firms will incorporate AI somehow, he added. Matthew Todd, a drug discovery professor at University College London, agreed that AI now threads through much of the field, labeling it a “catchall phrase” due to its varied applications.

AI is clearly reshaping drug development. Han cited initiatives at firms like AstraZeneca, Novo Nordisk, and GSK, noting AI can already propose drug concepts by suggesting molecules that might engage known disease-related targets or existing drug sites. Todd described it as highly effective for accelerating studies and “road testing” concepts. Given Anthropic’s frontier model expertise, it would likely leverage generative AI to explore extensive chemical and biological spaces, uncovering links that humans might miss, which could yield fresh drug ideas, new targets, or repurposed therapies.

Yet this remains far from any AI-created treatment reaching patients. Todd stated the sector is “a long way off” from regulatory approval of an AI-designed drug for human use. He stressed that the process would stay human-guided throughout. Both experts highlighted insufficient high-quality public data on chemical behavior in the body as a potential bottleneck, noting persistent knowledge gaps even in well-researched biology areas.

AI models “haven’t yet come close to making experiments unnecessary.”

AI cannot resolve many bottlenecks in drug discovery. Frank von Delft, an Oxford professor of structural chemical biology and protein crystallography head at the Oxford Centre for Medicines Discovery, acknowledged excitement around AI progress but cautioned that models “haven’t yet come close to making experiments unnecessary.” Candidates must still undergo real-world checks for effectiveness, safety, and practical traits for preparation, storage, and delivery. These steps demand expertise, funding, and time, particularly human trials where many candidates fail. If Anthropic pursues a drug, von Delft said, it “is going to have to spend a lot on experiments.”

Anthropic may be prepared to invest. Over the past year it has recruited biologists and established wet labs, with multiple life sciences openings active. Han observed active recruitment, noting several academic peers received offers. Without specifics, he believes Anthropic has drawn talent from major pharma and top universities.

Given these layers, any chosen disease area means returns are distant, likely spanning nearly a decade due to typical trial timelines. There is “always a big lag time” in medicine testing, Todd said. “It takes time to show experimentally that something’s safe.” No AI-designed drug has completed trials and gained FDA clearance for market. Some candidates have reached trials, yet AI’s exact role and performance edge remain unclear. AI can hasten discovery phases, but medicines must still succeed via traditional, deliberate real-world testing.

Follow topics and authors from this story to see more like this in your personalized homepage feed and to receive email updates.

Robert Hart