Page 339 - CyberSecurityEssentialsEnglish

Research integrity policies usually define misconduct through three central categories:
fabrication, the creation of data or results that do not exist; falsification, the manipulation
or omission of materials, processes, or data in a way that distorts the record; and plagiarism,

the use of another’s ideas, words, or results without proper attribution [11]. Other harmful
practices exist that do not meet the same formal definition, such as selective reporting,

honorary authorship, or poor data management.

AI can create new routes to old problems. It may generate data or quotations that look real,
rephrase a source text in a way that conceals its origin, fill a method section with details that
never happened, or write a discussion attributing an interpretation to authors without

evidence. The problem is not that a machine produced the text, but that the research record
does not represent what happened, or does not attribute knowledge to those who produced
it.

Publication ethics guidance states that AI tools do not meet the conditions of authorship and

may not be listed as authors [14]. Editorial guidance affirms the same: human authors are
responsible for accuracy, citations, and originality, and the use of assisting technologies must
be disclosed according to journal policy [12][13]. The details of disclosure may differ

between bodies, so the instructions of the journal, the funder, or the university should be
reviewed before submission.

Authorship is not a reward for status or funding alone. It requires an intellectual contribution

and responsibility for the work according to the standards of the field. Contributor role
taxonomies help describe roles such as conceptualization, methodology, data curation,
analysis, software, visualization, original drafting, and review and editing [15]. But they do
not on their own determine who deserves the title of author; they clarify contributions and

reduce ambiguity.

The use of AI should be discussed within the team early. Which tools are permitted? Which
data must not be uploaded? Who reviews the outputs? How is the use documented? Who

approves the final version? An early agreement reduces disputes and prevents a member
from using a tool in a way that participants’ consent or institutional policy does not allow.
Good disclosure depends on the significance of the use. Simple language editing may be

mentioned according to the required policy. But using a tool in designing a query, extracting
data, writing code, coding interviews, creating a figure, or interpreting results needs a
clearer description in the method or an appendix: the name of the tool, the version or date

of access, the purpose, the nature of the inputs, and how human beings verified the outputs
[17].

Do not upload protected text or confidential material merely because the purpose is
academic. The constraints may include copyright, data agreements, ethical approvals, and

338