Anthropic Says It Discovered a CRISPR-Like System. Now What?

"The experiments are still in the queue. The PR is already live," says one expert.

By Emily Mullin and Anna Rogers | Science | September 29, 2026

Anthropic announced this week that its AI models have uncovered a biological mechanism it describes as "CRISPR-like"—a claim that, if validated, could mark a significant milestone in AI-driven scientific discovery. But the announcement has also reignited a familiar debate: how much credit is due to a model before the science is actually confirmed in the lab?

What Anthropic Claims to Have Found

According to the company, its models identified a previously unknown protein-based system that appears to function similarly to CRISPR—the gene-editing technology that revolutionized molecular biology after its discovery in bacteria. Where CRISPR uses RNA-guided proteins to target specific DNA sequences, Anthropic says its system operates through a distinct mechanism the company believes could be adapted for targeted genomic applications.

The findings were presented alongside computational analysis and structural predictions, but Anthropic has not yet published peer-reviewed experimental confirmation.

The Gap Between Prediction and Proof

The announcement drew immediate skepticism from researchers who noted that AI-generated hypotheses, however elegant, still require wet-lab validation.

"The experiments are still in the queue. The PR is already live," one expert told WIRED, capturing the tension between the pace of AI-driven announcements and the slower rhythm of bench science.

This is not a new dynamic in 2026. Over the past two years, AI labs have increasingly released biological findings—new antibiotics, protein binders, and metabolic pathways—ahead of independent replication. Each case raises the same questions: What counts as a discovery? And who gets credit for it?

Why It Matters

If validated, an AI-designed CRISPR-like system would represent more than a technical curiosity:

• Gene editing: A new mechanism could offer alternatives where CRISPR has limitations, such as delivery, off-target effects, or size constraints.
• Biosecurity: A novel system with gene-editing potential also carries dual-use concerns, and regulators are still developing frameworks for AI-discovered biological tools.
• Scientific credit: Attribution questions loom large when a model—not a researcher—proposes the mechanism.

What Happens Next

Anthropic says it is working with external laboratories to test the system, though it has not disclosed timelines or partners. Until those experiments conclude, the claim remains exactly what it is: a compelling computational prediction awaiting empirical proof.

For now, the more instructive story may be about process. AI is compressing the timeline between hypothesis and headline—but not, so far, between headline and evidence.

Reporting by Emily Mullin and Anna Rogers for WIRED Science.

via Wired AI