Anthropic: Claude Designs Scorpion Toxin Binders

Anthropic says its AI model Claude designed small proteins that attach to scorpion neurotoxins, with 17 of 211 lab-tested designs meeting its criteria for a binder. In a research post published on 20 October, the company said it had “asked an internal research model to design minibinders (small proteins that attach tightly to a target) against scorpion neurotoxins.”

The work is an early test of whether AI can speed up drug discovery for neglected diseases, but it stops well short of a medicine. Anthropic is explicit: “We have not confirmed whether these binders neutralize scorpion venom in a cellular environment, so we do not claim to have produced an antivenom.” The results are Anthropic’s own, and its post cites no peer-reviewed paper.

In brief

Anthropic says Claude generated 318,484 protein designs over a campaign of about 64 hours; 211 were lab-tested and 17 met its binder criteria.
Each of the 17 binds one of four scorpion neurotoxins known to be dangerous to humans, according to the company.
Anthropic has not tested the designs in cells or animals, claims no antivenom, and is releasing all 211 designs and their data.

A campaign run mainly by the model

According to Anthropic, “Over the next 64 hours, an internal research model, operating within Claude Science, worked mainly autonomously,” with staff directing it through four follow-up prompts. The prompts were developed with Citrini Research, which Anthropic describes as “the financial investment research firm”. The project follows the company’s August minibinder study.

Claude used 12 design tools to generate 318,484 designs and co-folded and screened 553,052 design-toxin pairs, the post says, using “on the order of 10,000 GPU hours”. It “selected 180 of them for testing,” and Anthropic’s team “added 31 more, with Claude’s help.” The company adds that “Under our supervision, Claude analyzed the research data, generated the figures, and drafted the final report.”

The targets were five neurotoxins from scorpions in four regions: Mexico (Cn2), Brazil (Ts1), North Africa (AaH I and AaH II) and the Middle East (LqhIII). Anthropic notes that “Scorpions sting 1.2 million people or more every year, resulting in more than 3,250 deaths, mostly among children”.

Seventeen binders, each against a single toxin

Two contract research organisations, Adaptyv Bio and Twist Bioscience, tested all 211 designs. To count as a binder, a design had to show binding in every replicate at both labs, with every dissociation constant (KD) of “1 µM or tighter”, and no signal against an unrelated human protein, according to the post.

Toxin	Binders meeting Anthropic’s criteria
Cn2 (Mexico)	7
AaH II (North Africa)	4
AaH I (North Africa)	4
LqhIII (Middle East)	2

Source: Anthropic research post, 20 October 2026.

None of the binders was against Ts1, and attempts to design binders against several toxins at once fell short. “None of the designs made against multiple toxins met our criteria for more than one,” Anthropic says. “Among the 60 single-toxin designs, 11 were binders (18%),” a rate it calls consistent with typical protein design campaigns. Its two strongest binders, Cn2-B1 and Cn2-B2, bound Cn2 with affinities “in the single-digit nanomolar range”, and competition experiments suggest they attach on or near the same site as an antibody called LR, the company says.

A first step, according to outside experts

Anthropic sent the work to two outside experts, Susana Vázquez Torres of the Spanish National Cancer Research Centre (CNIO) in Madrid and Timothy P. Jenkins of the Technical University of Denmark, “who said they felt this work was a promising first step.” Vázquez Torres, quoted in the post, said “binding itself is no longer the main bottleneck” and added: “The challenge is to use these extraordinary tools to develop molecules that can become real drugs, ones that are safe, affordable, and widely accessible.”

The limits are substantial. Anthropic says binding was shown only against toxins grown in the lab rather than purified from venom, and that “The two biggest limitations of this study are that we didn’t test any design for neutralization in vitro (against cells grown in the lab) or for efficacy in vivo (in living animals, such as mice).” It is releasing all 211 designs and their data and has published a technical report.

Our read

This is TSN’s analysis. The interesting result is the speed and autonomy of the design campaign, not a new therapy: the company’s own caveats rule out any claim about treatment. The 18% hit rate on single-toxin designs is, by Anthropic’s account, typical rather than exceptional. The open data release is what lets outside labs test whether these binders hold up.