Anthropic Claude Biology Breakthrough Faces Questions Over Enzyme Research

Anthropic’s claim that its Claude AI independently identified a previously uncharacterized enzyme system is facing scrutiny after a computational biologist said his team had already been studying the same enzymes for several years.

Mario Rodríguez Mestre said he and his colleagues had spent roughly four years researching ART enzymes and related molecules. He also said the team had shared important findings with Anthropic while using the company’s AI models.

That has raised a central question about Anthropic’s reported breakthrough: how independently did Claude arrive at the discovery, and how much information may have come from research previously shared with the AI system?

The New York Times reported the dispute on Sunday, citing Mestre’s concerns about the origins of the discovery.

Mestre first identified the enzymes in jumbo bacteriophages in 2022 while researching reverse transcriptases. He later briefly worked as a consultant for ReNegade Therapeutics. The company filed a patent in 2023 covering several reverse transcriptases, including enzymes Mestre had discovered, with Mestre listed as a co-inventor, according to the report.

“So, for me, the most important question is not ‘Were ARTs already known?’ They were,” Mestre told the publication.

Anthropic did not immediately respond to Benzinga’s request for comment.

Claude’s CRISPR-Like Enzyme Discovery

Anthropic announced Wednesday that Claude had identified a previously uncharacterized enzyme system in bacteriophages after analyzing a massive DNA database.

According to the company, nearly 950 Claude agents processed approximately 210 million tokens over a 21-hour period. During that analysis, one of the agents identified an unusual repeating pattern that was then brought to human researchers for further investigation.

Scientists provided high-level guidance during the process and later conducted laboratory experiments to investigate the finding.

Anthropic described the enzyme system as having properties that resemble CRISPR, although its precise biological function remains under investigation. The company said it chose to share the early results because they demonstrate how Claude could potentially assist researchers with complex scientific problems.

Feng Zhang, one of the pioneers of CRISPR genome-editing technology, described the finding as “genuinely intriguing” and said it deserves further investigation.

The debate surrounding Mestre’s earlier research does not necessarily determine whether Claude’s analysis was technically useful. Instead, it raises a broader question about how AI-assisted scientific discoveries should be evaluated when models have been exposed to information from researchers working in the same field.

That distinction could become increasingly important as AI systems are used to search scientific literature, analyze proprietary datasets, generate hypotheses, and identify patterns that human researchers may have overlooked.

Anthropic Is Expanding Its Life Sciences Research

The enzyme discovery comes as Anthropic increases its focus on applying AI to biology and scientific research.

The company has argued that advanced AI systems could accelerate parts of the scientific discovery process, potentially helping researchers analyze large datasets, develop hypotheses, and streamline complex research workflows.

Earlier this month, Anthropic established a physical wet laboratory in the San Francisco Bay Area, expanding its life sciences work beyond computational research and into hands-on experimentation.

Eric Kauderer-Abrams, Anthropic’s life sciences leader, said laboratory work remains an important part of biological research. The company is combining its own research efforts with collaborations involving outside organizations.

Anthropic has clarified that the new facility is not specifically dedicated to drug discovery.

Anthropic and Novo Nordisk Partner on AI Research

Anthropic is also working with pharmaceutical companies as it expands its presence in life sciences.

Earlier this month, the company announced a collaboration with Novo Nordisk to explore ways AI could accelerate the development of new medicines.

Under the partnership, Novo plans to use Anthropic’s advanced AI models to improve software development and expand the use of AI across its operations. The collaboration includes an emphasis on data governance and human oversight as the companies explore how the technology can be deployed responsibly.

The partnership illustrates how AI companies are increasingly moving beyond general-purpose chatbots and into specialized scientific and pharmaceutical applications.

Claude Science Targets Scientific Research

Anthropic’s broader life sciences strategy also includes dedicated software for researchers.

In July, the company launched Claude Science, an AI workbench designed specifically for scientific users. The platform represents a significant expansion of Claude’s role in areas such as biology, chemistry, and other research disciplines.

Claude Science supports fields including:

• Genomics
• Proteomics
• Structural biology
• Cheminformatics

The platform also integrates more than 60 scientific databases and tools, giving researchers access to specialized information and computational resources within their workflows.

Anthropic has additionally connected the platform with Nvidia’s BioNeMo Agent Toolkit, providing access to specialized biological AI models and tools.

What the Dispute Means for AI-Driven Scientific Discovery

The controversy surrounding Claude’s enzyme discovery highlights a challenge that is likely to become more common as AI becomes more deeply embedded in scientific research.

An AI system may identify a pattern that appears new, but determining whether the underlying discovery is genuinely novel requires careful examination of existing research, unpublished findings, patents, datasets, and information that may have been provided to the model.

This becomes particularly complicated when researchers interact with AI systems while working on unpublished projects. Information shared during those interactions can raise difficult questions about attribution, provenance, confidentiality, and the boundaries of AI-assisted discovery.

In Claude’s case, Anthropic presented the enzyme system as a significant example of what its AI agents can accomplish when analyzing enormous biological datasets. Mestre’s account, meanwhile, raises questions about how much prior knowledge surrounded the enzymes before Claude’s analysis.

Those questions will ultimately need to be separated from the technical results themselves. Further scientific investigation will be necessary to establish what the enzyme system does, how novel the findings are, and what role Claude played in reaching them.

As Anthropic continues expanding into biology, drug research, and scientific computing, the episode illustrates both the potential and the challenges of using AI as a research partner. The technology may be capable of finding patterns across datasets that are difficult for humans to examine manually, but establishing genuine scientific novelty still requires rigorous validation, transparent attribution, and independent research.