FOURWEEKMBA EXPLAINER · FROM THE NEWSROOM · UPDATED 14 NOVEMBER 2026
The short answer: AI cannot be an author of a scientific paper. arXiv, Springer Nature, the Science journals, the ICMJE medical editors and the publication ethics body COPE all say so. The reason is the same everywhere: an author must take responsibility for the work, and a tool cannot. What researchers can do is say what the AI did, in the methods section or the acknowledgements. Physicist Edward Witten’s paper of 3 November 2026 shows how that looks in practice. It credits Anthropic’s Claude Opus 5.5 by name for a specific idea in Section 3, and thanks it in the acknowledgements, while Witten remains the only author.
Free to share with a link to this page. Page counts are our arithmetic from the paper’s contents page.
The paper is “A Note On Lattice Definition of Gauge Theories With Chiral Fermions” (arXiv:2611.05051). Edward Witten of the Institute for Advanced Study is its only author. It was posted on 3 November 2026 and is listed at 33 pages. It is a preprint: we have not seen a peer-reviewed version.
The physics problem is old. The abstract calls it “a longstanding challenge” to put four-dimensional gauge theories with chiral fermions on a lattice. It credits a recent algorithm to Craig and says the same variant was “also proposed recently by Hsu.”
Claude Opus 5.5 appears three times, by our count. Twice in Section 3, “Continuum Interpolations of Admissible Lattice Gauge Fields”:
“Anthropic’s Claude Opus 5.5 suggested a likely variant that gives a smooth interpolation with link holonomies that equal precisely the corresponding lattice link variables and such that the interpolated gauge field in a hypercube depend only on the link variables on the boundary of that hypercube.”
“For example, Opus 5.5 also described a variant in which the inductive step is made by a sort of linear interpolation rather than by solving a Dirichlet problem.”
And once in the acknowledgements, next to two human colleagues:
“I thank T. Hartman for pointing out Craig’s paper, and K. Yonekura for helpful comments on the manuscript. Assistance on many details was provided by Claude Opus 5.5.”
Three things stand out. The credit is specific: it names the idea the model suggested. Witten then writes the construction out himself, step by step. And the model sits where non-author help usually goes, in the text and the acknowledgements, not in the author line.
What the rules say
Every policy we read draws the same line: AI can help, but it cannot be an author. They differ on where to disclose the help.
arXiv
Generative AI tools “should not be listed as an author.”
Authors must report “any significant use of sophisticated tools”, now including text-to-text generative AI, “consistent with subject standards for methodology.”
Each author takes “full responsibility for all its contents, irrespective of how the contents were generated.”
Springer Nature (Nature journals)
“Human accountability is non-transferable.”
“Assigning authorship or accountability to AI systems or tools” is not permitted.
Uses are sorted into three risk levels, with disclosure expected.
Science journals
AI tools “may not be listed as authors or coauthors.”
Authors note AI use “in the cover letter and in the methods section or acknowledgments section.”
“Editors may decline to move forward with manuscripts if AI is used inappropriately.”
ICMJE (medical journals)
AI tools “should not be listed as authors because they cannot be responsible for the accuracy, integrity, and originality of the work.”
Disclose at submission, in the cover letter and the paper. Nondisclosure “may be construed as misconduct in some circumstances.”
COPE (publication ethics)
“AI tools cannot meet the requirements for authorship as they cannot take responsibility for the submitted work.”
Say how the tool was used, and which tool, in the methods section or similar.
Help with words, or help with ideas?
Springer Nature’s policy is the most detailed. It sorts AI use by risk, not by tool:
Green: permitted. Polishing language, structuring a manuscript. The test: the AI “does not introduce new intellectual content.”
Amber: permitted with oversight and disclosure. Examples include “Suggesting analytical, experimental or methodological approaches.” Here the AI “does introduce new intellectual content and requires verification and oversight.”
Red: not permitted. Examples include giving AI authorship, and “Generating hypotheses, analyses or conclusions and presenting them as human-derived.”
As we read it, a model suggesting a variant of a construction is new intellectual content, the amber kind. Witten’s paper was posted on arXiv, not in a Springer Nature journal, so the framework does not formally apply. But it shows what the paper does: it names the contribution, keeps it in the open and keeps the human in charge of checking it.
How to credit AI in a paper
Never in the author line. Every policy above rules that out.
Name the tool. COPE asks which tool was used. Witten names the model and version: Claude Opus 5.5.
Say what it did. Language help and new ideas are different. If the model suggested an idea, say so where the idea appears, as Section 3 does.
Put it where the journal asks. Methods, acknowledgements, cover letter, or all three. arXiv points to the norms of each field.
Check it yourself. Authors carry full responsibility for AI output, errors included, under arXiv’s policy and the others.
What is not established
The paper does not say how Witten used Claude beyond these three passages, how many sessions were involved, or which details the acknowledgement covers. It presents the suggestion as a “likely variant.” We have not checked the physics. The paper is a preprint, and we do not know where, or whether, it will be submitted for publication.
FAQ
Can an AI be listed as an author on a scientific paper?
No. arXiv, Springer Nature, the Science journals, ICMJE and COPE all say AI tools cannot be authors, because authors must take responsibility for the work.
Where do you disclose AI use in a paper?
Usually in the methods section or the acknowledgements, and often the cover letter. The Science journals ask for the cover letter plus the methods or acknowledgements. ICMJE asks for disclosure at submission.
How did Edward Witten credit Claude?
In Section 3 he wrote that “Anthropic’s Claude Opus 5.5 suggested a likely variant” of a construction, and that Opus 5.5 “also described a variant.” His acknowledgements say: “Assistance on many details was provided by Claude Opus 5.5.”
Is using AI for research ideas allowed?
Springer Nature allows AI to suggest methods or approaches, with human oversight, verification and disclosure. Presenting AI-generated hypotheses or conclusions as human work is not allowed.
Who is responsible if the AI makes a mistake?
The authors. arXiv says errors or misleading content from AI tools are “the responsibility of the author(s).”
Sources: Edward Witten, “A Note On Lattice Definition of Gauge Theories With Chiral Fermions”, arXiv:2611.05051v1, 3 November 2026; arXiv moderation policy, “Policy for authors’ use of generative AI language tools”; Springer Nature / Nature Portfolio editorial policy, “Artificial Intelligence (AI)”; Science journals editorial policies; ICMJE Recommendations, “Use of AI by Authors”; COPE position, “Authorship and AI tools”; FourWeekMBA newsroom report linked above. All read 14 November 2026. Quotes are verbatim from the sources named; the paper quotes are from the arXiv PDF only. The count of mentions and the page counts are ours.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
