A five-year-old girl stung by a scorpion at her Sonora, Mexico preschool died after her public hospital, which had exhausted its antivenom days earlier, located a fresh supply only 45 minutes after she arrived.
That gap between medical need and available treatment frames an investigation by Hunterbrook, which spent two days inside Anthropic's offices using an unreleased, less-restricted version of Claude to design 211 candidate molecules against scorpion toxins. The AI produced early binders for four of five targets but did not achieve the project's most ambitious goal, a single molecule capable of binding multiple toxins. Hunterbrook, which states that neither it nor its collaborator Citrini was paid by Anthropic for the work, concludes that AI can accelerate early-stage design without resolving the physical, clinical, manufacturing, and regulatory bottlenecks that determine whether a medicine ever reaches a patient.
Ticker: Not applicable (Anthropic is privately held; the report cites a reported pre-Thanksgiving IPO target)
Research Firm: Hunterbrook
Report URL: https://hntrbrk.com/investigations/antivenom?ref=shortreport.fyi
Position Disclosure: The report states that neither Hunterbrook nor Citrini was paid by Anthropic for the work. No short or other financial position is disclosed in the source report.
Thesis
Hunterbrook, working with Citrini and citing an Anthropic technical report reviewed by independent academic experts, tested whether frontier AI can meaningfully accelerate drug discovery beyond early molecular design.
- Mixed Binder Results: The AI-selected candidates produced initial binders against four of five targeted scorpion toxins, with the fifth yielding "a messier result," per Anthropic's technical report.
- Unproven Antivenom Function: The experiments "did not establish that these binders neutralize toxins or will work as an antivenom," and broader testing suggested some binders may be "sticky," leaving selectivity unconfirmed.
- Moonshot Goal Missed: Despite 211 AI-generated designs, Claude "did not generate a binder" meeting the criteria for a single molecule binding multiple toxins, the project's most ambitious technical objective.
- No Edge Over Specialists: The report alleges "a skilled team" could have produced comparable results with existing protein-design tools, and that Claude "did not break new ground specialists couldn't have with available tools."
- Unproven Efficiency Gain: The author argues AI likely made design faster, cheaper, and more accessible given a higher hit rate, but characterizes the broader conclusion as "hypothesized, but not yet proven."
- Bottleneck Unmoved: After the AI design phase, the project stalled for weeks on sourcing scorpion toxins and reagents, contracts, and shipping, consistent with former FDA leader Dr. Amy Abernethy's comment that AI "moves the choke point to clinical development."
- Years to Clinic, Even Funded: Comparable AI-enabled biotechs needed years to reach human trials: Triana took seven years and more than $100 million to enter a Phase I/II trial, while Superluminal, backed with $120 million including Nvidia's investment, planned clinic entry five years after its founding.
- Internal AI Gap: Anthropic's public model routes bioscience work to a less capable system, while the unreleased model used in this experiment ran faster, was unconstrained by compute, and relied on prompting tactics reporters were not permitted to record.
Catalysts
- IPO timeline scrutiny: Anthropic's reported IPO target before the U.S. Thanksgiving holiday, per Bloomberg Law, would subject the company's AI-biotech narrative to market scrutiny.
- Preclinical validation milestones: Further preclinical testing to confirm binder selectivity and toxin neutralization is the next data point needed to move from design to a proven antivenom candidate.
- Development-stage progression: Any progression of the AI-designed candidates into animal studies, manufacturing, or regulatory review would signal whether the lab result can survive drug development's standard gauntlet.
- Supply-chain pressure signals: Continued antivenom supply shortages, including Southern Africa's main supplier remaining out of stock until the end of 2026, keep real-world urgency separate from the AI timeline.
- Future Anthropic biomedical output: Any further publication from Anthropic's biomedical research lab or its reported biotech acquisition, first reported by The Star on September 18, 2026, would clarify what additional resources are meant to deliver.
Company Response
The source report does not describe a separate request for comment sent to Anthropic. Anthropic's position is instead represented through its direct participation, hosting the two-day experiment, co-funding the CRO testing, and publishing its own technical report on the results, which the authors say was reviewed by independent academic experts. The report does not describe Anthropic responding to the broader critique that AI does not resolve the clinical, manufacturing, and regulatory bottlenecks downstream of molecular design.
Notable Details
- Southern Africa's main scorpion-antivenom supplier had run out of stock by October, with no new supply expected until the end of 2026, according to expert Jonathan Leeming.
- Traditional scorpion antivenom is produced by immunizing horses with venom and harvesting the resulting antibodies, a process Leeming called "100 years old"; the AI-designed candidates were produced without horses.
- Mexican scientists spent 25 years developing the antivenom behind Anascorp, the sole FDA-approved U.S. scorpion antivenom, followed by roughly another decade of U.S. clinical trials before approval.
- Report author Nathaniel Horwitz previously co-founded two AI-enabled biotech ventures himself, Triana (2019) and Superluminal (2021), and says his own first experimental medicine failed.
"So while software continues to devour the world, it must also gnaw on the gristly bones of medical development and actually getting medicines to patients across the world."
The author writes this after cataloguing the clinical, manufacturing, regulatory, insurance, and access barriers that remain once AI finishes designing a molecule.
FAQs
Does Anthropic have a publicly traded stock or ticker?
No. Anthropic is privately held. The report cites Bloomberg Law reporting that Anthropic was targeting an IPO before the U.S. Thanksgiving holiday, but no ticker currently exists for public trading.
What is Anthropic's role in this scorpion antivenom experiment?
Anthropic hosted a two-day visit from Hunterbrook and Citrini at its New York City office, provided an unreleased, less-restricted version of Claude, and co-funded two contract research organizations that synthesized and tested 211 AI-generated molecular designs. It later published a technical report on the results, which the authors say was reviewed by independent academic experts.
Who is Hunterbrook and what did it conclude about Anthropic's AI?
Hunterbrook is the investigative newsroom that conducted the experiment alongside Citrini; the report states that neither organization was paid by Anthropic. Hunterbrook concluded that the AI accelerated early molecular design but did not achieve the project's most ambitious goal and did not resolve the physical, clinical, and regulatory bottlenecks that follow design work.
What is scorpion antivenom and why is it in short supply?
Antivenom is the only treatment able to directly neutralize circulating venom toxins, according to emergency physician Dr. Jesús Chávez Guzmán. More than one million people are stung by scorpions annually, with thousands of deaths, mostly children, per a cited PMC publication; single suppliers dominate production in both Mexico and Southern Africa, creating shortage risk when supply runs low.
Did Anthropic's AI actually create a working antivenom?
No. The AI-designed binders showed initial activity against four of five targeted toxins, but the report states the experiments "did not establish that these binders neutralize toxins or will work as an antivenom." Broader testing suggested some binders may be "sticky," leaving selectivity unconfirmed.
What is Anascorp and how does it relate to the AI experiment?
Anascorp is the sole FDA-approved scorpion antivenom in the United States, made from horse-derived antibodies. Mexican scientists spent 25 years developing it, followed by roughly a decade of U.S. clinical trials before FDA approval, a timeline the report uses to illustrate how much further development AI-designed candidates would likely need.
Why did Southern Africa's antivenom supply run out?
Expert Jonathan Leeming told Hunterbrook that the region's main supplier had run out of antivenom by October, with no new stock expected until the end of 2026, highlighting a single-supplier dependency separate from any AI-related development.
What is Hunterbrook's overall view on AI and drug development?
Hunterbrook argues that AI, including Anthropic's unreleased Claude model, is an efficiency improvement for early molecular design rather than a near-term breakthrough that can rapidly produce deployable medicines. The report points to the slow lab, clinical, manufacturing, and regulatory stages that followed the AI design phase as evidence the core bottlenecks remain unchanged.
Disclaimer: This summary is not primary research and does not constitute investment advice. It is a brief overview of a detailed equity research report authored by the firm, organization, or source referenced in this article or at https://hntrbrk.com/investigations/antivenom, which contains extensive evidence, regulatory filings, and analysis; readers are encouraged to review the full report there for a comprehensive understanding. The content provided in this publication is not authored or originated by us — we act solely as a distributor and do not endorse, verify, or take responsibility for the accuracy, completeness, or reliability of the information presented. This publication is for informational purposes only and should not be construed as legal, business, investment, or tax advice. Always conduct independent due diligence and consult qualified professionals before making any decisions based on the information contained herein. We disclaim all liability for any loss or damage arising from reliance on third-party content, and the views expressed are solely those of the respective source and do not necessarily reflect our own.