AI

Anthropic Research Demonstrates Claude-Designed Scorpion-Toxin Binders

Tuesday, October 20, 2026Read Original

Details

  • Anthropic says Claude generated more than 300,000 protein designs in about 64 hours to target five scorpion neurotoxins.
  • The project was conducted with Citrini Research, with Adaptyv Bio and Twist Bioscience testing the candidates in laboratory assays.
  • Of 211 designs submitted for testing, 17 met Anthropic’s criteria for binding to dangerous toxins.
  • The successful binders targeted toxins associated with scorpions from Mexico, North Africa, and the Middle East; none met the criteria for the Brazilian target.
  • The designs are minibinders, or small proteins intended to attach to a toxin’s cell-binding region and potentially block its activity.
  • Anthropic said the work does not constitute an antivenom because cellular neutralization, animal studies, and further validation remain outstanding.
  • The company released all 211 tested designs and the accompanying laboratory data for outside researchers.

Impact

The result supports a shift toward AI systems handling parts of the protein-design workflow, from candidate generation to experimental prioritization. It also highlights the current boundary of such systems: producing binders is materially earlier than demonstrating toxin neutralization or clinical utility. By releasing the tested designs and data, Anthropic may help researchers reproduce the work and compare Claude’s approach with emerging AI protein-design platforms, while leaving the costly biological validation phase unresolved.

Rift Dispatch