Anthropic Builds Secret Biological Wet Lab, Capable of Conducting Real Biochemical Experiments
According to an exclusive report by Reuters, AI company Anthropic has quietly established a wet lab in the San Francisco Bay Area, equipped with the physical conditions necessary for conducting real biochemical experiments, marking the company's official expansion into drug development, previously focused solely on software.
Eric Kauderer-Abrams, head of life sciences at Anthropic, confirmed the lab's existence, stating, "To do biological research, the ultimate test still lies in real lab work." Some experiments are conducted in-house while others are completed in collaboration with external partners, consistent with practices of most biotech companies. He also revealed that life sciences is one of the largest areas of current human and resource investment for Anthropic, although the company is still in the "very early" stages of automating lab operations. Notably, an Anthropic spokesperson stated that the lab is "not specifically for drug development," but declined to elaborate on its specific uses, scale, staffing, activation time, and biosafety level.
In recent months, Anthropic has been active in the life sciences sector: acquiring startup Coefficient Bio for approximately $400 million in stock to build drug development tools; Novartis CEO Vas Narasimhan joining the company's board; launching software named "Claude Science"; publicly announcing the start of a drug development project in June in San Francisco; and recruiting for positions related to procurement operations and experts in protein and nucleic acid characterization on LinkedIn, explicitly stating the goal of "increasing the pace of progress in life sciences by an order of magnitude."
Anthropic is targeting the "undruggable" rare disease field that traditional pharmaceutical companies have sidelined due to high target difficulty and limited commercial returns, believing that AI excels at handling complex molecules like bispecific and even trispecific antibodies that require simultaneous targeting of multiple sites and are challenging to design. CEO Dario Amodei has stated that his motivation for entering this field is personal, as his father passed away from a disease for which a cure was only developed years later. Currently, Anthropic has clearly defined its boundaries: it only conducts preclinical research, does not engage in clinical trials, and does not directly compete with pharmaceutical companies—this is also to alleviate concerns from large pharmaceutical clients (including Genentech, Bristol Myers Squibb, and Novo Nordisk) about Anthropic potentially learning from their data.
These actions come as Anthropic prepares for an IPO with an estimated valuation of around $2 trillion, coinciding with ongoing controversies over AI safety: on September 9, former pre-training researcher Jacob Coxon announced his resignation and warned that the AI lab is "heading straight towards self-improving superintelligence, betting human lives on it." Two current Anthropic employees also confirmed similar concerns within the company; around September 10, a threat report released by Anthropic disclosed that the company had intercepted five attempts by researchers to use Claude for dangerous dual-use biological research, one of which involved personnel from a military background attempting to draft a grant application for research on enhancing the function of the chikungunya virus, which was intercepted and reported to law enforcement. The report also acknowledged for the first time that the risk of its newer models "assisting in biological weapon development is no longer definitively below the safety threshold." The same report revealed that a group with Russian ties had used Claude Code to build a drone swarm system capable of autonomously locking onto "human" targets and issuing detonation commands without human authorization, which had been tested in related operations in Yemen.
From a funding and business logic perspective, Anthropic's intensive layout in life sciences is essentially telling the story of a second growth curve for an upcoming massive IPO, stating that "AI is more than just chatbots; it can penetrate tangible scientific discovery": acquiring drug development tool capabilities through the purchase of Coefficient Bio, bringing in industry resources like Novartis CEO Narasimhan for endorsement, while strictly adhering to the boundary of "only doing preclinical, not touching clinical" to reassure paying pharmaceutical clients like Genentech, Bristol Myers Squibb, and Novo Nordisk about concerns over data being "stolen." However, at the same time, the company's own researchers are intensively issuing warnings about AI out of control and biological weapon risks, creating a clear tension with its increased focus on biological lab operations—on one hand, telling the capital market the story of "AI accelerating new drug development" to boost valuation, while on the other hand, needing to prove to regulators and the public that its AI systems are safe. This coexistence of "accelerating" and "pulling the handbrake" will objectively affect investors, partner pharmaceutical companies, and regulatory agencies' judgments on its IPO valuation and business expansion speed.
Source: Public Information
ABAB AI Insight
Anthropic's approach of oscillating between AI safety narratives and business expansion is not new—it has previously released "threat reports" disclosing cases of its systems being misused while simultaneously launching more powerful models (such as the five cases disclosed this time involving chikungunya virus gene function enhancement and autonomous drone attack systems). This pattern of "expanding while self-disclosing risks" is essentially a continuation of its long-standing "responsible expansion" public relations and compliance strategy, rather than an isolated incident.
In terms of funding pathways, Anthropic's acquisition of Coefficient Bio for approximately $400 million in stock (rather than cash) indicates that this transaction did not deplete the company's cash reserves but instead used its own valuation as "currency" to acquire drug development tool capabilities, a common low cash cost expansion path for high-valuation tech companies. Bringing in Novartis CEO Narasimhan to the board is a way to exchange industry connections and credibility for pharmaceutical client trust, adding narrative chips for the upcoming approximately $2 trillion valuation IPO that "AI can penetrate tangible science."
This layout can be compared to Google's Isomorphic Labs, which has been advancing AI drug discovery for years—initially planning to enter clinical stages by the end of 2026, but has already delayed once, reflecting the industry reality that it generally takes years from molecular discovery to drug market launch, and most candidate drugs fail in clinical stages. In terms of the industry's current stage, AI drug development is in the early transition from "computational simulation" to "physical experimental validation," and Anthropic's establishment of a wet lab is a typical action during this transition period, with significant commercialization challenges ahead.
Structural judgment: This event belongs to the reconstruction of the industry chain. By building its own wet lab, acquiring drug development tool companies, and bringing pharmaceutical executives onto its board, Anthropic is attempting to extend its AI capabilities from merely being a "software tool supplier" to the preclinical research phase of the drug discovery industry chain, which has been entirely dominated by traditional pharmaceutical and biotech companies, while maintaining collaboration space with pharmaceutical companies through the boundary of "only doing preclinical, not touching clinical"—essentially inserting a new "AI preclinical discovery" phase into the drug development industry chain, redefining the division of labor between AI companies and the traditional pharmaceutical industry chain.
ABAB News · Cognitive Law
- Expanding the lab while shouting about uncontrolled risks.
- Valuation stories need microscopic evidence.
- Undruggable diseases have become opportunities for AI.