Anthropic Announces "Self-Developing New Drugs," Experts Pour Cold Water: FDA Approval Will Take at Least 10 More Years

AI can not only write code, but now it also wants to invent life-saving drugs! According to a report by authoritative tech media The Verge today (3rd), Anthropic, the developer of the renowned AI model Claude, dropped a bombshell, announcing that in addition to launching an AI workbench dedicated to scientific research, the company will officially enter the physical pharmaceutical field, focusing on developing cures for 'neglected diseases'. This means Anthropic will transform from a pure software provider into a potential competitor of large pharmaceutical companies. However, experts have also poured cold water: AI drugs still need to endure at least 10 years of clinical trials before they can truly pass FDA certification and reach patients. (Prerequisite: Anthropic spent $100k on a hackathon, using real scientific research data to test Claude Science) (Background: Challenging Nvidia's dominance! Anthropic reportedly teams up with Samsung to develop custom AI chips, Claude's computing infrastructure will undergo a major overhaul)

Table of Contents

Toggle

• From selling shovels to mining yourself, Anthropic ventures into physical pharmaceuticals
• AI Drug Development: Tech Companies' Ambitions vs. Medical Reality's Harshness
• Expert: AI Cannot Replace Lab Rats in the Laboratory

At the crossroads of accelerating integration of artificial intelligence and life sciences, San Francisco-based AI giant Anthropic has chosen the most difficult but also the most ambitious path.

This week, at Anthropic's 'The Briefing: AI for Science' event, the company first showcased the brand new **Claude Science ** — an 'AI workbench' designed for top scientists. It integrates scattered research tools and massive datasets into a single environment and automatically generates complex visual charts, aiming to significantly accelerate the process of scientific discovery and medical intervention.

However, what truly shocked the industry was the heavy news in the second half of the event.

From selling shovels to mining yourself, Anthropic ventures into physical pharmaceuticals

Anthropic's life sciences head Eric Kauderer-Abrams announced at the event that the company will not just stop at the stage of 'providing AI tools to pharmaceutical companies', but will dive in to develop drugs themselves. The initial R&D focus will be on 'neglected diseases' that lack commercial profit and are rarely pursued by traditional pharmaceutical companies.

This decision makes Anthropic particularly unique in the tech circle. Although OpenAI, Google, and Amazon have all launched life science-related tools, very few frontier AI model developers dare to publicly announce that they will develop physical drugs themselves. This means that in the future, Anthropic will on one hand sell Claude to large pharmaceutical companies, and on the other hand, may become their direct competitor in specific disease areas.

AI Drug Development: Tech Companies' Ambitions vs. Medical Reality's Harshness

Despite Silicon Valley's optimism about AI disrupting the pharmaceutical industry, many interviewed top scholars and experts have stated that transforming computational models into real-world life-saving drugs still faces a huge reality gap.

| Observation Dimension | | --- | Current Status and Layout of Tech Industry | Realistic Warnings from Medical Experts | | --- | --- | --- | | Changes in R&D Model | Anthropic actively recruits biologists and has established proprietary physical 'wet labs'. | AI excels at accelerating the early 'search' for candidate molecules, but the entire drug development process still relies on strict human expert supervision and physical validation. | | Clinical Trials and Regulations | Several startups (e.g., Insilico) have AI-assisted designed candidate drugs entering clinical trial stages. | Currently, 'not a single' drug fully designed by AI has successfully passed complete clinical trials and received FDA approval for market. | | Time and Data Bottlenecks | Expectation to generate perfect drug formulas with one click via large language models (LLMs). | Lack of high-quality public experimental data; and from discovering a new target to final market launch, it still takes at least 10 years of long time. |

Expert: AI Cannot Replace Lab Rats in the Laboratory

Oxford University professor Frank von Delft bluntly pointed out that AI is still far from replacing physical experiments. Even if AI can perfectly predict the structure of a new molecule, this candidate drug must still undergo tests for toxicity, stability, and in vivo efficacy in the real world. These preclinical and clinical stage tests not only cost hundreds of millions of dollars but also have a terrifyingly high failure rate.

University College London drug discovery professor Matthew Todd also added that although AI has great potential in quickly testing new drug ideas, the chemical reactions within living organisms are extremely complex, and many medical fields still have huge knowledge gaps that cannot be filled by algorithms alone.

Currently, Anthropic has not disclosed the specific disease target of its first candidate drug, nor has it explained whether it plans to 'go it alone' in the clinical trial and manufacturing stages or seek cooperation with traditional pharmaceutical companies. But there is no denying that this AI giant, valued at tens of billions of dollars, has taken the most challenging step toward its grand life sciences blueprint.