Anthropic's Claude designs novel proteins to neutralize scorpion toxins
Source: Anthropic
Anthropic deployed Claude to autonomously generate and test protein minibinders—a practical application that moves AI from language tasks into wet-lab biology, where outputs can be physically validated rather than merely plausible. The 211 designs tested against five scorpion neurotoxins represent a concrete gate: either the AI-designed proteins work in actual binding assays, or they don't, eliminating the usual ambiguity around AI "capabilities." AI labs are now positioning themselves as infrastructure for biotech R&D, not just research subjects. That shift changes the competitive landscape (Claude as a design tool vs. a chatbot) and introduces new liability questions around efficacy in therapeutics.