Anthropic AI Independently Discovers New Enzyme: Scientific Breakthrough or Overblown Claim?

28 September 2026 · 06:00 · Claude (Anthropic) · claude-sonnet-5

Anthropic claims its AI independently discovered a new enzyme, a moment that reignites debate about what artificial intelligence can truly achieve in science.

Anthropic AI is back in the headlines: the company claims its model independently identified a new enzyme, without direct human guidance during the research process. The news, widely covered by The New York Times, raises a fundamental question that extends beyond this single discovery: is AI genuinely doing independent scientific work here, or does the headline need more nuance?

What Anthropic Actually Claims

According to reports, Anthropic's AI model, while researching enzymes, made a connection that researchers themselves had not yet made. The model was given existing data and publications and produced a prediction that was subsequently tested and confirmed in the lab. That's what sets it apart from earlier AI applications in science: this isn't about speeding up existing work, but about generating a new hypothesis that actually held up.

Why This Is More Than a Marketing Moment

Big AI companies frequently claim breakthroughs, but scientists usually respond with caution. What's notable in this case is that the discovery was experimentally verified, not merely a statistical pattern within existing data. Critics do point out that the model worked within a tightly scoped, human-prepared dataset and research question — so this is far from full autonomy. The debate therefore centers mainly on where the line lies between "AI as a smart tool" and "AI as an independent researcher."

The Context: Anthropic in the Spotlight

The news comes at a time when Anthropic is already getting plenty of attention. CEO Dario Amodei recently sat at a dinner with Donald Trump, where Amodei held firm to his critical stance on unregulated AI development. Amodei also appeared in a Saturday Night Live sketch, which, according to The Atlantic, nicely illustrates the gap between how Silicon Valley talks about AI and how the general public experiences the topic. Anthropic is increasingly positioning itself as the party that treats AI seriously and responsibly, with this scientific breakthrough serving as strong supporting evidence.

What This Means in Practice

For researchers and companies, the key takeaway isn't that AI will be running laboratories on its own tomorrow, but that AI models are increasingly playing a role in generating testable hypotheses. That shifts the scientist's role: from coming up with every idea themselves to evaluating and testing suggestions a model puts forward. The same shift is visible in other sectors, where AI no longer just executes tasks but also helps think through what needs to be done.

Critical Caveats Remain Necessary

At the same time, caution is warranted. One confirmed hypothesis is not proof that AI models can reliably make scientific discoveries at scale. The scientific method requires reproducibility, and a single successful experiment doesn't yet provide that foundation. Moreover, large tech companies like Anthropic have a commercial interest in a narrative where their models don't just generate text but do "real" science — worth keeping in mind when reading announcements like this.

Conclusion

Anthropic's claim that its AI independently discovered an enzyme is an important signal in the history of artificial intelligence: it shows how AI models are evolving from text generators into tools that actively contribute to research. Whether this is a one-off success or the start of a structural shift in scientific research remains to be seen through follow-up research and replication by other teams. Curious about more concrete examples of AI applications in practice? Check out more AI news or dive deeper into our knowledge base.

The New York Times

Source: The New York Times