Anthropic claims AI-assisted discovery of a CRISPR-like enzyme system in bacteriophages
The AI lab reports its biology team, supported by Claude, identified a novel enzyme system with properties similar to CRISPR in just 21 hours of model effort.
1 source · single source
- Anthropic disclosed a biology lab discovery of a CRISPR-like enzyme system in bacteriophages, found with AI assistance from Claude.
- The discovery was made within months of opening the lab, with Claude contributing about 21 hours of effort using 950 agents and 210 million tokens.
- Anthropic emphasized human-conducted lab work; fully autonomous wet-lab experiments remain future work.
- CEO Dario Amodei noted prior related work by Stanford researchers and acknowledged risks of bioterrorism alongside potential medical benefits.
Anthropic announced that its newly established Bay Area biology lab has identified a previously unknown enzyme system in bacteriophages with properties reminiscent of CRISPR, a widely used gene-editing technology derived from bacterial immune systems.
The company said the discovery was made "mostly, though not entirely, by Claude," with the model contributing about 21 hours of concerted effort across roughly 950 agents that processed approximately 210 million tokens.
Anthropic described the enzyme system as capable of performing operations like cutting, copying, and pasting DNA, though the broader research community will need to validate the novelty and significance of the finding.
Anthropic CEO Dario Amodei noted that a Stanford team had previously reported a system in some ways similar to the one found by Claude, underscoring that the discovery builds on prior work.
The lab, which operates at biosafety levels BSL-1 and BSL-2 and does not handle human-infecting pathogens, conducts all physical experiments with human scientists, according to the company.
Amodei also reiterated concerns about bioterrorism risks even as he expressed optimism that AI could help cure most diseases within five to ten years.
He added that fully autonomous wet-lab experiments—where Claude would control lab equipment—remain a future possibility rather than current practice.
- Oct 13, 2026 · Google DeepMind — Blog
Google DeepMind details safety guardrails for Gemini 4 Argon’s cyberdefense and biology capabilities
Trust76 - Oct 13, 2026 · Google DeepMind — Blog
Google DeepMind unveils Gemini 4 Argon for robotics with 87% score on Embodied Reasoning benchmark
Trust78 - Oct 13, 2026 · arXiv cs.CL
Researchers propose Elastic Latent Recurrence to unify recurrent and non-recurrent language models without inference-time trade-offs
Trust79