Independent newsroom

Wednesday, October 21, 2026

The File

Every claim traces to a primary source. Uncertainty is labeled. Corrections are public.

Anthropic · Wednesday, October 21, 2026

Anthropic says Claude designed proteins that grab scorpion toxins

The File desk · Oct. 21, 2026, 6:40 a.m. ET

Partially confirmed1 primary sources

Status

Partially confirmed. Anthropic posted the research on Oct. 20, 2026, and this desk re-opened the page. The binding results are the company's account of tests at two contract labs. They are not peer-reviewed, and Anthropic says it has not tested whether any protein neutralizes venom.
  • Confirmed
    Anthropic said an internal research version of Claude designed small proteins aimed at scorpion nerve toxins, and that two labs tested 211 designs.
  • Partially confirmed
    Anthropic says 17 designs met its own bar for a binder. No outside journal has reviewed the work, and the company says it has not tested neutralization.
Anthropic said Tuesday that an internal version of Claude designed small proteins that stick to scorpion nerve toxins, a possible early step toward a new antivenom. Two outside labs tested 211 designs. Anthropic says 17 met its bar for a binder, and it has not tested whether any of them stop the venom.1
Adaptyv Bio and Twist Bioscience tested the designs. Anthropic's bar was binding in every replicate at both labs, a dissociation constant of 1 micromolar or tighter, and no signal against an unrelated human protein. The best two, which it calls Cn2-B1 and Cn2-B2, bound a Mexican scorpion toxin in the single-digit nanomolar range. None of the designs aimed at more than one toxin met the bar for more than one. The targets were five toxins from Mexico, Brazil, North Africa, and the Middle East.1
The research post is dated Oct. 20. Anthropic says scorpions sting 1.2 million people or more every year and that more than 3,250 people die, mostly children in places with limited medical care. Those figures are Anthropic's, attributed on its page. The company says Claude used 12 design tools, generated 318,484 designs, and, 64 hours after the first prompt, produced a list of 180 designs. Staff added 31 more, with the model's help, for 211 that were tested. The campaign used on the order of 10,000 GPU hours. The model “worked mainly autonomously,” with four follow-ups. The prompts were developed with Citrini Research.1
Why it matters (Estimate). Estimate. Scorpion stings kill thousands of people a year, mostly children, and today's antivenoms are slow and costly to make from horse blood. If a lab can produce a credible starting point for a neglected disease in days, more of those illnesses might get research at all. This is not a treatment. Anthropic says it has not shown that any of these proteins work against real venom.1
Anthropic says, “We have not confirmed whether these binders neutralize scorpion venom,” and that it did not test any design in cells or in animals. Binding was shown against lab-grown toxins, not venom milked from a scorpion. The company is releasing all 211 designs and the data. Timothy P. Jenkins of the Technical University of Denmark, quoted on Anthropic's page, called designing binders against five toxins in a matter of days “a great showcase of the AI-driven paradigm shift we are seeing in therapeutic discovery.”
How we know

Notes

  1. 1.Anthropic Science, “How Claude designed protein binders against scorpion neurotoxins,” published Oct. 20, 2026. Re-opened for this edition. https://www.anthropic.com/research/claude-scorpion-toxin-binder-design “Teams at Adaptyv Bio and Twist Bioscience experimentally tested 211 designs for binding. Of these, 17 met our criteria.” “Scorpions sting 1.2 million people or more every year, resulting in more than 3,250 deaths.” “Claude used 12 design tools to generate 318,484 designs.” “Sixty-four hours after we gave Claude our initial prompt, it produced its final list of 180 designs. The campaign used on the order of 10,000 GPU hours.” Staff added 31 designs. “worked mainly autonomously, although we directed its work with four follow-ups.” Prompts with Citrini Research. Criteria include KD “1 µM or tighter.” Cn2-B1 and Cn2-B2 bound Cn2 “in the single-digit nanomolar range.” “None of the designs made against multiple toxins met our criteria for more than one.” “We have not confirmed whether these binders neutralize scorpion venom.” “we didn't test any design for neutralization in vitro … or for efficacy in vivo.” Timothy P. Jenkins, Technical University of Denmark: “Designing credible binders against five toxins in parallel, in a matter of days, is a great showcase of the AI-driven paradigm shift we are seeing in therapeutic discovery.” Susana Vázquez Torres, CNIO Madrid, is also quoted on the page. The technical report and the Hugging Face dataset were not opened. Company X post, 12:18 p.m. Eastern Tuesday: https://x.com/AnthropicAI/status/2112579251555721573

What is still unknown or disputed

  • Whether any binder neutralizes venom or protects an animal. Anthropic says that was not tested.
  • The work is not peer-reviewed.
  • The internal research model is not named.
  • The technical report PDF and the public dataset were not opened.

Primary sources

Every claim in this story is drawn from the documents below. If a fetch failed, that is recorded on the card.

  1. Source 1

    How Claude designed protein binders against scorpion neurotoxins

    Anthropic · October 20, 2026

    Teams at Adaptyv Bio and Twist Bioscience experimentally tested 211 designs for binding. Of these, 17 met our criteria. We have not confirmed whether these binders neutralize scorpion venom.

    https://www.anthropic.com/research/claude-scorpion-toxin-binder-design

Ask this story

Answers come only from this story and its sources. Free test: limited questions per day.

0/300