The October 8 development in the dispute over Anthropic’s September 23 ART announcement is Mario Rodríguez Mestre’s account of overlapping research: he says he shared unpublished work through a private Claude account, but cannot demonstrate that Anthropic used it.
Mestre’s account of unpublished research
Mestre, a computational biology PhD candidate at the University of Copenhagen, says his unpublished research described the same viral signatures as the reported system. He says he shared dissertation drafts, analyses and related material through a private Claude account.
Anthropic has denied that its biology team accessed user transcripts or that Claude was trained on them. NeoTeo previously covered Mestre’s questions about attribution.
What Anthropic says Claude found in ART
On September 23, Anthropic announced that Claude agents had helped identify array-associated reverse transcriptases (ARTs) in bacteriophages, viruses that infect bacteria. The reported system combines a reverse transcriptase, a neighboring partner gene and an array of regularly spaced DNA repeats. The reverse transcriptase itself had been identified before.
Anthropic says its initial experiments found that the repeat array produces distinct short RNAs. The system’s biological function remains unknown. The work was a preprint and had not been peer reviewed as of October 8.
Anthropic says roughly 950 agents spent 21 hours searching, using 210 million tokens. They reviewed more than 200,000 reverse transcriptases, selected 3,500 candidate systems and narrowed those to 20 for closer analysis. Anthropic says human scientists carried out all the laboratory work.
What the additional searches indicate
Ten additional search runs reportedly missed the repeat array, indicating that those runs did not recover the pattern.
