Abstract
Corporate use of bots and AI to handle enquiries is no longer unusual. Problems arise, however, when a company does not disclose that a response is automated, attaches the full name and job title of an actual employee, deliberately presents the response as individual handling by a human staff member, and then treats that response as evidence that the enquiry has already been answered. This article examines an operation observed at Springer Nature. It differs from an ordinary bot with a human-like name and simultaneously undermines AI transparency, the authenticity of complaint handling, the accuracy of staff identification, and the accountability expected of a listed company. From December 2025 to January 2026, an automated solicitation bot bearing the name of Managing Editor Fei Tan continued after explicit refusals and requests that the messages stop. In September 2026, responses bearing the name of Assistant Editor Alfeya Baig were sent after Article 50 of the EU AI Act had become applicable. Improper handling contrary to Article 50 was therefore observed in a research-integrity matter.
1. Overview
At Springer Nature, automated responses displaying the names and job titles of actual people were observed in communications attributed to an Assistant Editor and a Managing Editor.
From December 2025 to January 2026, repeated messages bearing the name of Fei Tan, Managing Editor, solicited a submission to a journal where improper peer review had occurred. The journal was Discover Artificial Intelligence, whose peer-review process COPE found improper in 2026. Because Fei Tan is an actual Managing Editor, the author did not initially identify the messages as bot output. The author requested action on the problems at the journal and asked that the solicitation of new submissions stop. The messages nevertheless continued, and a “Last reminder” was sent even after the author explicitly wrote “STOP HARASSMENT.”
From August to September 2026, responses bearing the name of Alfeya Baig, Assistant Editor, repeatedly ignored the author's specific questions and instead asked the author to provide more information about the matter. Each new ticket received only one automated reply. No further response followed even when the author complied with the bot's instruction. When the author raised this problem with Springer Nature employees in customer support and elsewhere, they treated the automated message as a response from a real human staff member and closed the enquiry as “answered” or “resolved.”
The author was therefore forced to resend the same questions and records, seek the reopening of wrongly closed enquiries, and repeatedly explain the history to human staff. Springer Nature's continued treatment of low-quality AI output as formal responses imposed many hours of additional work. Multiple automated responses using the names and job titles of actual people were observed, although their legal status is not identical.
2. Characteristics of the AI Bots Observed after Article 50 Became Applicable
- The channel is an official journal contact point.
- Each ticket receives only one response.
- The system cannot retain the submission ID, personal submission information, or questions that it cites in its own message.
- It gives irrelevant responses such as “please tell us more about this case” or asks whether the enquiry concerns an article-processing-charge discount, repeatedly demanding information already supplied.
- The sender is a shared journal mailbox rather than an individual's email account.
- The displayed name is the full name of an actual person and is accompanied by an actual corporate job title such as Assistant Editor.
- When the matter was reported to headquarters, human employees treated the existence of the automated reply as proof that a response had already been sent and refused further action because the matter was supposedly resolved.
| Comparison | Messages Bearing Fei Tan's Name | Messages Bearing Alfeya Baig's Name |
| Use | Submission enquiries, invitations to special issues involving article processing charges, and enquiries about discounts on article processing charges | Individual enquiries, handling of submitted manuscripts, and complaints concerning research integrity |
| Display | Fei Tan, Managing Editor | Alfeya Baig, Assistant Editor |
| Content | Primarily promotional | An individual response to specific questions sent by the author |
| Interactivity | Strongly resembles an automated email campaign | Responds to an individual user's enquiry |
| AI disclosure | None | None |
| Appearance of a real person | Full name of a real person plus the title Managing Editor | Full name of a real person plus the title Assistant Editor |
| Period | Mainly December 2025 to April 2026 | August to September 2026 |
| Article 50 of the EU AI Act | Before Article 50 became applicable; the principal issues concern misleading representation and unfair commercial practices | After Article 50 became applicable. The AI answered individual enquiries directly, and headquarters employees treated its output as an official answer from the editorial office. The transparency obligation for AI systems intended for direct interaction therefore applies. |
| Central issue | Automated solicitations were presented as personal communications from a real Managing Editor and continued after requests that they stop. | The AI handled an individual matter under the identity of a real Assistant Editor without disclosure, and human employees then treated its output as a formal human response. |
3. Difference from Ordinary Bots with Human-Like Names
Other companies also give bots human-like names. In many such operations, however, a boundary remains to prevent confusion with human staff. The companies concerned are not identified here.
The seriousness of this case does not arise from the display alone. Springer Nature created the appearance of a human staff member through a real name and job title, sent inaccurate bot output, and later treated that output as proof that it had already answered the enquiry. This gave the bot output the professional credibility of a real person and the institutional effect of an official response.
| Comparison | Ordinary Human-Named Bot | Springer Nature Operation Observed in This Case |
| Name | First name only, nickname, or no surname | Full name of an actual staff member |
| Job title | No employee job title is assigned to the bot | An actual corporate job title is displayed |
| Automation disclosure | Often identified as an automated reply or AI assistant | No indication that the response comes from a bot or automated system |
| Appearance of human judgment | Bot output is distinguished from a human's formal decision | The response appears to show that the named staff member personally reviewed the matter |
| Treatment after an incorrect response | An incorrect bot response is not treated as proof that the matter is resolved | The bot's request for more information was later used by headquarters employees as the basis for saying that the matter had already been resolved |
4. Article 50 of the EU AI Act
Article 50(1) of the EU AI Act requires providers of AI systems intended to interact directly with natural persons, as a general rule, to design and develop those systems so that users are informed that they are interacting with an AI system. An exception applies where this is obvious to a reasonably well-informed, observant and circumspect natural person. The transparency obligations in Article 50 have applied since 2 August 2026.
The European Commission's 2026 guidelines state that the disclosure should be clear and distinguishable from the first interaction. Its purpose is to allow users to understand the nature of the system with which they are interacting and to adjust their trust accordingly. Springer Nature's customer-support operation did the opposite. It concealed that the response was generated by AI and displayed the full name and job title of an actual person, actively creating the appearance that the user was communicating with that employee.
Article 50(1) applies directly where the mechanism is an AI system within the meaning of the Act and the AI itself responds directly to a natural person. Because this case involved case-specific emails generated by generative or conversational AI, the September 2026 responses were issued after Article 50 became applicable and violated its transparency requirement.
5. Section 5 of the German UWG: Complaint Handling, Identity, Qualifications and Status
Springer Nature has substantial operations in Germany as well as the United Kingdom and is listed on the Frankfurt Stock Exchange. Section 5 of the German Act against Unfair Competition (UWG) expressly includes the provision of services, after-sales customer assistance, complaint handling, and a trader's identity, qualifications, status and affiliation among the matters about which a commercial practice may be misleading.
In complaint handling, it therefore matters who answered and whether a human Assistant Editor actually reviewed the case. Presenting a bot response under the name and position of a human staff member and treating it as formal handling may mislead the recipient about the nature of the complaint process.
6. Conflict with Springer Nature's Published AI-Automation Policy
In its response to UTIE Instruments Inc., Springer Nature acknowledged that some journal emails soliciting submissions and responding to enquiries, including the emails discussed here, were automated using an AI bot.
Automation by bots is not a business policy that Springer Nature conceals. In annual reports and official announcements, the company actively presents investment in AI and automation as a business achievement. Its 2024 annual report stated that more than 90 active AI-enabled processes and solutions were operating across the research-publishing process by the end of that year and explained that AI was strengthening business operations.
Under its “human-centred AI initiatives,” Springer Nature states as a transparency principle that it discloses when AI systems are being used and as an accountability principle that it maintains human oversight of the development and outputs of its AI tools and solutions.
Its 2025 annual report likewise describes continued large-scale investment in AI to process more content and improve the customer experience. Automation of production and corporate processes remains a priority investment area, and the company reports investment in workflows from submission to publication, peer-review support, and automated peer-review processes. An official announcement in March 2026 presented the use of approximately 60 AI tools across more than 1.5 million papers in 2025 as an achievement.
Chief Executive Officer Frank Vrancken Peeters also stated that AI would reduce the burden on authors and simplify workflows for editors and reviewers while maintaining:
transparency about where AI is used, with clear human oversight and accountability
| Springer Nature's Published AI-Automation Policy | AI-Automation Operation Observed |
| Actively expand the use of AI | AI use was expanded without identifying the system as a bot |
| Publicise automation as an investment and business achievement | The automated nature of the response was concealed while the full name of a real person was displayed |
| Disclose when AI is used | Actual employee titles such as Assistant Editor and Managing Editor were displayed |
| Maintain human oversight and accountability | The system returned context-free responses such as requests for more information or offers to discuss article-processing-charge support, did not respond further, and repeated the same response when a new email was sent |
| Ensure transparency about AI use | Human employees at headquarters treated the automated output as a formal research-integrity response from a human staff member and used it as grounds for marking the matter answered or resolved |
7. The Bot's Incorrect Output Was Used as a Human Response
The seriousness of this case is not determined merely by the bot repeating the same question. The bot replied only once to each enquiry and did not function as continuing support. By displaying the same real person's name and job title, however, it created the appearance that one Assistant Editor remained responsible for the matter and that a continuing conversation was taking place. This also differs from an ordinary one-time automated reply.
If the bot ended each enquiry by replying with another question, it should have disclosed at that point that the automated process was complete. The gap between appearance and reality became greater when Springer Nature employees later treated the inaccurate one-off response as a handling record showing that the journal had already answered or that the editorial office had replied. Springer Nature's operation was deceptive because it caused the user to believe that support remained active. The combination of an actual person's name and job title, the absence of bot disclosure, and the later treatment of the incorrect bot output as a formal response made the practice particularly serious.
The enquiries concerned AI-generated peer review, a self-investigation by the handling editor, and a fictitious investigation reported by the Research Integrity Team to COPE. The author was therefore asking whether a human expert had actually conducted an assessment or investigation. The very channel used to verify that question falsely appeared to be operated by a human employee while the matter was in fact being processed by an automated bot.
8. Conclusion
A company that introduces AI support must not mislead users into believing that they are receiving human support. Even if a human-like name is assigned, displaying the full name and professional title of an actual employee and making it appear that this individual personally reviewed the matter is deceptive and clearly contrary to Article 50 of the EU AI Act, which has applied since August 2026.
In research-integrity cases, who reviewed the matter, who made the decision, and who continued to handle the case are facts central to the authenticity of the procedure. Presenting a bot that terminates after one reply as continuing support from a real Assistant Editor, and using its context-free output as grounds for saying that the matter was resolved, undermined not only transparency but the authenticity of Springer Nature's complaint handling. Section 5 of the German UWG also expressly regulates misleading representations concerning complaint handling, identity, qualifications and status.
References
- Regulation (EU) 2024/1689 (AI Act), Article 50 and Article 113. EUR-Lex
- European Commission, Guidelines on transparency obligations for providers and deployers of AI systems (20 July 2026). European Commission
- European Commission, Transparency obligations under Article 50 of the AI Act - Q&A. European Commission
- German Act against Unfair Competition (UWG), Section 5. Federal Ministry of Justice
Primary records: Special-issue solicitation emails bearing the name of Fei Tan, Managing Editor, and the author's replies dated 19 and 20 December 2025 and 16 and 30 January 2026; emails bearing the name of Alfeya Baig, Assistant Editor at Discover Artificial Intelligence, dated 19 August and 3 September 2026; and the author's replies and journal-support forwarding records.