分享好友 Global News首页 Global News分类 切换频道

Fake AI Papers Falsely Name Hinton, Fei-Fei Li, LeCun and Andrew Ng as Authors

2026-11-16 11:5910关注AI滥用的WeChat Official Account

Alan Said, a professor of computer science at the University of Gothenburg, received an alert from Google Scholar: He had a new paper published.

The paper is titled "Intelligence Systems Supporting Sustainable Global Development" and was published in a journal called American International Journal of Research and Innovation (AIJRI). The title is related to his research direction of "AI and Sustainability", and the signature is indeed his.

The problem is: he didn't write this paper.

“It would not be surprising if I had written a paper in this area, but the paper and its references make it look AI-generated. It has no content; it is just a bunch of slop,” Said told Times Higher Education (THE).

Following this paper, THE uncovered a bigger problem: a group of little-known journals were publishing junk papers generated by AI in batches in the name of professors from top universities such as Stanford and Berkeley. Those "signed" include Geoffrey Hinton, Yann LeCun, Yoshua Bengio, Fei-Fei Li, Christopher Manning, Michael I. Jordan, Andrew Ng, Stuart Russell and many other well-known researchers in the AI ​​field.

There are even junk AI papers that pile the names of many famous researchers together.

The paper remained online as its byline changed

Said's first reaction was to complain. He contacted the journal's domain name registrar, who responded that the article had been removed.

But the paper didn’t disappear, the author just changed from Said to someone else. Two days later, the author was changed again, according to screenshots Said provided to THE. The same paper was published under the names of three different people.

This detail is quite telling: for this kind of journal, the content of the paper is not important at all, but the name on it is important. The name can be replaced at any time, just like changing a model picture on the product page.

Said subsequently discovered two "sibling journals": American Journal of Advanced Multidisciplinary Research and Innovation (AJAMRI) and American Journal For Multidisciplinary Research (AJFMR). The three journals have similar visual styles and list the same editor-in-chief and editorial-board members.

Among them, AJAMRI has published dozens of papers signed by well-known scholars, and the list can be called the "all-star lineup" in the AI ​​circle: Stanford's Fei-Fei Li and Christopher Manning, and Berkeley's Michael I. Jordan and Stuart Russell.

Stanford University computer science professor Jure Leskovec is also on the list. He made his position clear to THE: "I can confirm that I did not write this paper, contributed to it, and did not authorize the use of my name. I was not even aware of its existence before receiving your email. This is quite worrying."

The domain name registered in 2026 has "back issues" from several years ago.

If you look carefully at these three journals, it is not difficult to find flaws.

All three journal websites use domains registered in 2026, yet display archives of papers that they claim were published years earlier. They uniformly call themselves "peer-reviewed, open access, multidisciplinary" journals, and clearly charge publication fees.

The charging rules are also worth a look. AIJRI and AJAMRI charge US$200 for papers with "five or more non-Indian authors"; AJFMR charges US$80 for one to four authors, and only US$10 for five or more authors.

The editors-in-chief of all three journals have the same name: Denish Brown. The contact details are a WhatsApp number with a US area code. The editorial board lists dozens of members, mostly from Indian universities.

The editorial-board listings also raised questions. Kanakaraju Manupati, a postdoctoral fellow at Virginia Commonwealth University, is among them. He told THE: "I didn't know the existence of this journal, and I have no connection with it. I just checked the journal's website and was surprised to see my name on the editorial board."

In other words, the "academic endorsement" of these journals, from the authors to the editorial board, is likely to be stolen in its entirety. THE contacted three journals but received no response as of press time.

Manning: "This is a worse scam"

After the report was issued, the named Christopher Manning responded on 𝕏 that he had seen two fake papers "signed" by him, which were published in the so-called International Journal of Computer Vision and AI Applications and International Journal of Clinical and Medical Sciences.

The former's name is obviously taking advantage of the popularity of the top computer vision journal IJCV (International Journal of Computer Vision); the latter is even more outrageous, as an NLP scholar "published" it in a clinical medical journal. It’s worth noting that these two journals were not among the three investigated by THE, meaning similar operations may involve far more than just one gang.

“Scam journals have always been around; this is an even worse scam,” Manning wrote. “The bottom line is don’t let them be indexed! Please, Google Scholar!”

A user complained in the comment area that Google Scholar has been recommending some obviously fake papers to him, and there are a lot of big names in the author column. “I really can’t figure out what the motive is. Why would someone do this?”

Manning’s reply points out the key: Manning suggested the journals may list prominent scholars to attract submissions from doctoral students; the journals charge publication fees. He finally added: "Maybe it will actually work!"

A misleading editorial-board listing

This is what Said is most worried about: “If a new PhD student sees my name in this journal, he might think it’s okay to apply here, and I really don’t want that to happen.”

Traditional predatory journals rely on "quick publication and guaranteed publication of submitted manuscripts" to attract authors who are eager to increase the number of papers. But they generally have one shortcoming: they don’t look like serious journals. Generative AI has just made up for this shortcoming. It can "write" papers in batches in a few minutes with standardized formats, complete citations, and seemingly meaningful papers. With the names of top scholars, a pheasant journal can "package" decent back issues and a luxurious lineup of authors in a few weeks.

This method of play is not the first time it has appeared. Diomidis Spinellis, a professor of software engineering at the Athens University of Economics and Business, encountered exactly the same thing in December 2023: a paper titled "Global Business Strategies in the Digital Age" was published in the Global International Journal of Innovative Research (GIJIR) under his name and unit, and was accompanied by a valid DOI.

Spinellis didn't stop at clarification. He crawled the entire GIJIR website for analysis, and published a case study in "Research Integrity and Peer Review" in 2025. Of the 53 articles with the fewest citations, at least 48 appeared to be AI-generated, with the Turnitin test giving AI probabilities as high as 100% on several papers. His conclusion is: On the one hand, such journals use fake AI papers to enhance their status, and on the other hand, they use this to attract authors who want to increase the number of papers.

According to a report by JMIR in June this year, Spinellis classified GIJIR's papers into three categories: the majority are "fake signed papers" that are generated by the journal itself using AI and impersonate the names of scholars; followed by "purchased signed papers" that the authors spend money to buy; genuine research is in the minority.

Similar events will obviously become more intensive in 2026. In April this year, well-known cardiologist Eric Topol discovered that his name was used fraudulently in a digital health journal; in May, NBC News reported on a network of fake journals that had published hundreds of AI-generated papers, some of which also used the names of professors from top universities.

The real vulnerability lies downstream

If these papers just sit on a website that no one cares about, the harm may be limited. But Manning's appeal points to the real crux of the problem: indexing.

Said discovered this paper precisely because Google Scholar included it on his homepage and actively pushed it to him. Once included in academic search engines, these papers will appear on the scholar's personal homepage, enter the search results, and even be cited by other papers. Spinellis’s fake paper received three citations within a few months of publication.

There is a more insidious risk for the AI ​​industry: a large amount of model training data comes from scraped academic literature. When junk papers generated by AI are labeled with the names of real scholars and entered into the index, they may flow back into the training corpus of the next generation model, forming a polluted closed loop of "AI generation, AI collection, and then AI learning."

Technically speaking, this is not an unsolvable problem. The directions proposed by Spinellis in his research include strengthening the identity verification mechanism of DOI and ORCID, improving AI content detection, and reforming the scientific research evaluation system based on the number of papers. The JMIR report also mentioned that researchers can use checklists such as Think.Check.Submit to check the ISSN, DOI and independent indexing status of the journal before submitting their manuscript.

But all these measures cannot avoid the reality: the cost of generating a fake paper has approached zero, and the cost of clarifying a fake paper falls entirely on the victim scholars. After Said complained, the paper was changed to another author and continued to be published; Leskovec did not know that he had "published" the paper until the reporter sent an email; Manupati was also listed on the editorial board without knowing it.

So, here comes the question: When names like Fei-Fei Li and Manning can be borrowed at will, to what extent can the academic signature itself serve as a certificate of trust? In an era when AI makes "writing papers" cheap, what the academic community needs to protect may not only be the authenticity of the content, but also the authenticity of the name.

点赞 0
举报
收藏 0
分享 1