Real DOI numbers carried authors who never lived into the academic archive
Researchers counted 1,655 documents on Zenodo attributed to recurring AI-generated names and carrying real DOI numbers, while Google Scholar and Semantic Scholar indexed such material without verification. Anthropic also opened 10,000 research seats and applications for up to 50,000 dollars in project credits. Wider access to AI research tools is arriving alongside a new provenance problem for scholarly infrastructure.
Artificial Intelligence··Morning
Ghost names receive real identifiers
A preprint by researchers at Samsung and the University of Warsaw searched academic publishing for names such as Elena Vasquez and Marcus Chen, which large language models repeatedly produce when asked for experts. The team found 1,655 documents attributed to these names on Zenodo, the repository operated by CERN. The documents carry real DataCite DOI numbers, yet cite journals that do not exist and use fabricated publication dates. Because anyone with a free Zenodo account can create a DOI, the formal identifier does not mean the content or author has been verified. The work is not presented as peer-reviewed.[1]
Indexes connect synthetic research groups
The same synthetic names form linked research groups on ResearchGate, while Google Scholar and Semantic Scholar index the material without verification. The researchers report that some model families lean toward particular names: Elena Vasquez, for example, stands out in Claude Sonnet 4 output. For now, that pattern offers a clue to which model family may have produced a document. Lead author Michał Brzozowskim cautions that the distinction may weaken as texts carrying these names spread online and enter later training data. The risk extends beyond one fabricated paper because an unverified document can become visible again through layers of scholarly search infrastructure.[1]
Anthropic widens research access
Anthropic opened 10,000 free and discounted Claude team seats to scientists worldwide for one year. Standard seats are free, while seats with 5 times the usage limit cost 15 dollars a month. Applicants must be a principal investigator or equivalent at an academic or nonprofit research institution, and accepted applicants can add researchers from their own labs. The company also expanded its AI for Science programme beyond biology and opened applications for up to 50,000 dollars in credits per project. Biology and chemistry researchers remain limited to Opus-class models, while Fable models decline professional biology and drug-development questions because of dual-use concerns.[2]
