Anthropic: Internal Claude Model Designed 17 Lab-Confirmed Binders Against Scorpion Neurotoxins; Partner Citrini Is Pitching Life-Sciences Stocks
Anthropic says an internal research model designed 318,484 protein binders against scorpion neurotoxins; 17 of 211 lab-tested designs met its binder criteria. It does not claim an antivenom. Quartz relays Citrini Research's account that prompting required a biosecurity-classifier exemption at Anthropic's offices; Citrini is now backing life-sciences stocks.
A frontier model ran most of a binder-design loop, and the outside partner is an investment firm pitching stocks on the result. Anthropic published on Tuesday, October 20 results from a campaign in which an internal research model, working inside Claude Science, designed small protein binders against scorpion neurotoxins. Over 64 hours, with four follow-up prompts developed with Citrini Research, Claude generated 318,484 designs and selected 180 for testing; Anthropic's team added 31 more. Contract labs Adaptyv Bio and Twist Bioscience tested all 211, and 17 met Anthropic's criteria for a binder — each against a single toxin. None of the multi-toxin designs met the criteria for more than one toxin.
Anthropic says it has not tested the binders for neutralisation in cells or animals and does not claim to have produced an antivenom. The campaign used on the order of 10,000 GPU hours. Anthropic is releasing all 211 designs and their data.
Partner conflict and the classifier account
The project was run with Citrini Research, a financial investment research firm. Quartz reports Citrini's account that prompting took place at Anthropic's offices because Citrini's request to bypass a biosecurity classifier required security measures it lacked. That detail is Citrini's, relayed by Quartz; Anthropic's research post does not mention a classifier exemption.
Quartz, citing Bloomberg, says Citrini is now backing life-sciences stocks including Charles River Laboratories, Revvity and Tecan off the result. Citrini's stock pitch is a conflict of interest for any framing of the science as market-moving.
What the lab found
Anthropic targeted five neurotoxins (Cn2, Ts1, AaH I, AaH II, LqhIII). The two strongest designs, both against Mexican scorpion toxin Cn2, bound in the single-digit nanomolar range and appear to compete with a known neutralising antibody (LR), Anthropic says. Binding was measured by SPR at both CROs under Anthropic's hit criteria (KD ≤ 1 µM, no off-target signal against CLEC12A). Outside experts Susana Vázquez Torres and Timothy P. Jenkins called the work a promising first step in Anthropic's post.
Limits
- No neutralisation or in vivo efficacy data; not an antivenom claim.
- Biosecurity-classifier detail is Citrini via Quartz, not Anthropic's post.
- Citrini is an investment firm pitching stocks on this result.