Generative Artificial Intelligence (GAI), particularly through a wide variety of technologies—most notably Large Language Models (LLMs)—is rapidly being integrated into academic research, teaching, and administrative processes. These tools offer researchers and educators significant efficiency gains through functions such as text generation, code writing, image generation, data summarization, and language translation.

However, the uncontrolled or uninformed use of GAI technologies can give rise to serious ethical and legal risks, such as scientific misrepresentation, plagiarism, data privacy violations, copyright infringement, and algorithmic bias. This situation necessitates a clear, actionable, and enforceable guidance framework at the institutional level.

This guide has been prepared to support Yıldız Technical University academic staff in using AI tools in an ethical, responsible, and legally compliant manner. The rules and recommendations in this guide have been developed based on the relevant TÜBİTAK guide, the TÜBİTAK AYEK Regulation, the Personal Data Protection Law No. 6698 (KVKK), and the relevant ethics guide of the Council of Higher Education (YÖK).

1. Purpose, Scope, and Basic Definitions of the Guide

1.1. Purpose and Scope

The primary purpose of this guide is to explain to the academic staff of Yıldız Technical University, through concrete examples, under what conditions and how generative AI tools can be used within the scope of educational and research activities; which uses are restricted or prohibited; and why and how the disclosure of such use is mandatory. This guide aims to safeguard research integrity, ensure fairness and transparency in evaluation processes, ensure compliance with legal obligations regarding the protection of personal data, and safeguard intellectual property rights.

This guide applies to researchers, faculty members, research assistants, and all academic staff conducting projects within the university. This guide may be used as a reference for project applications submitted to funding agencies, including TÜBİTAK; in peer review and panel evaluation tasks; in the preparation of course materials; and in scientific publishing activities.

1.2. Basic Definitions

2. Fundamental Principles Underpinning the Guide

The rules and recommendations in this guide are based on the following fundamental principles. All provisions of the guide must be interpreted and applied within the framework of these principles:

1. Responsibility and Accountability: No matter how advanced AI tools may be, humans always bear ultimate responsibility for the content produced and the decisions made. AI systems cannot be held legally or ethically accountable.
2. Transparency: The use of AI in support processes and publishing activities must be clearly disclosed. Transparency both supports accountability and strengthens institutional trust in these processes.
3. Fairness and Non-Discrimination: Measures must be taken to mitigate potential algorithmic biases in AI tools. The use of AI must not provide an unfair advantage to specific individuals or groups.
4. Privacy and Data Protection: Application proposals, evaluation reports, and research data are confidential. The provisions of the Personal Data Protection Law (KVKK) must be followed when processing personal data; sensitive information must not be entered into AI tools under any circumstances.
5. Research Integrity and Originality: The use of AI must be conducted in a manner that does not lead to acts of scientific misconduct, such as plagiarism, data fabrication, or data manipulation.
6. Human-Centered Approach and Oversight: AI systems should be used to support human capabilities; they should not eliminate the human role in decision-making processes. Meaningful human oversight must always be maintained in critical processes.

3. Permitted Uses and Concrete Examples

AI tools may play a supportive role for the following purposes. A disclosure requirement applies to all of these uses (see Section 5). Academic staff will not accept outputs generated by AI systems at face value; they will evaluate the content with a critical eye and verify its accuracy using independent sources. Intellectual responsibility for the final content rests with the user in all cases.

3.1. Language and Style Improvement

Grammar and spelling checks, sentence structure improvements, strengthening of academic style, and foreign language support are considered within this scope. The AI system can be used to enhance the presentation of the text without altering its meaning or original content. However, before these processes are carried out, personal data and third-party intellectual property content must be removed from the input text.

3.2. Literature Review and Summarization

AI can be used to understand general trends in a specific research field, develop research questions, and grasp the main ideas of large text collections. However, none of the references suggested by the AI system (particularly author, year, journal, and title information) should be used under any circumstances without being independently verified in reliable academic databases such as Google Scholar, PubMed, or Web of Science. The tendency of the AI system to generate nonexistent articles as if they were real (hallucination) poses the highest risk in this field.

3.3. Coding and Technical Support

ÜYZ can be used to generate data analysis code for the research process, to correct errors in existing code, and to explain the code’s functionality. Additionally, ÜYZ can be tasked with coding solutions planned by the researcher and featuring task separation at the atomic level. The researcher must meticulously review the functional accuracy, efficiency, and potential licensing issues of the generated code. In this regard, having the researcher define test scenarios and expected results and providing these criteria to the ÜYZ as guidelines will enhance the reliability of the process.

4. Uses Requiring Caution

The following uses involve situations where the AI plays a more central role and therefore carries a higher risk. While the use of AI systems in these areas is not prohibited, the researcher is required to conduct comprehensive validation, clearly demonstrate their original intellectual contribution, and fully disclose its use.

4.1. Creating a Project Proposal or Draft Section of a Paper

Using AI tools to generate draft text for the introduction, methods, literature review, and results sections is a practice that is becoming more common but carries serious risks. The generated texts are often superficial and generic; they may fail to reflect the project’s unique context and may contain fabricated references, unrealistic promises, and factual errors. Using such drafts without sufficient revision by the researcher can pose serious risks.

4.2. Preparing Instructional Materials and Exam Questions

ÜYZ can be used to create a course presentation outline, a case study scenario, or an exam question. However, the content accuracy, alignment with course learning outcomes, difficulty balance, and originality of the generated materials must be verified by the instructor under all circumstances. The possibility that exam questions generated by ÜYZ may overlap with existing online sources should not be overlooked.

5. Prohibited Uses

The forms of use listed below are strictly prohibited as they clearly violate ethical principles, research integrity, confidentiality obligations, and applicable laws and regulations.

5.1. Plagiarism

It is strictly prohibited to use ÜYZ tools to copy, paraphrase, or present ideas, texts, data, methods, or visual materials belonging to others as one’s own work without citing the source or obtaining permission. Because content generated by AI systems may verbatim reproduce copyrighted works found in educational data, researchers are advised to check AI outputs using plagiarism detection software.

5.2. Fabrication and Falsification of Data

It is strictly prohibited to generate data via the AI system for experiments, measurements, or analyses that were not actually conducted, or to alter existing data or research results. These actions are defined as serious ethical violations.

5.3. Entering Confidential, Personal, or Unpublished Data into AI Tools

It is strictly prohibited to enter or upload confidential information related to project applications, unpublished research findings, personal data protected under the Personal Data Protection Law (student information, patient data, participant identification information, etc.), or information constituting trade secrets belonging to third parties into AI tools. Data uploaded to publicly accessible third-party ÜYZ platforms may be stored, processed, or used for model training by service providers. This violates both confidentiality obligations and the provisions of the KVKK regarding data processing and cross-border transfers (Article 9).

6. Disclosure of AI Tool Usage

In accordance with the principles of transparency and accountability, it is mandatory to disclose all AI tool usage that goes beyond basic spelling and grammar checks. This obligation reaffirms the researcher’s responsibility for the accuracy and originality of the content, ensures that reviewers and readers are aware of how the content was produced, and enables the university to monitor trends in AI tool usage.

When making a declaration, the exact version of the AI tool used and the date of use must be specified (e.g., Claude Sonnet 4.6 – March 23, 2026). The reason for specifying the date is that, even if AI models share the same version code, their learned weights change due to continuous learning.

6.1. Disclosure Threshold

The table below provides examples to clarify which uses require a declaration and which do not:

6.2. Declaration Method

The declaration can be made in the following ways, depending on the context of use:

• In TÜBİTAK applications: The tool used, its version, and the scope of its use must be entered in the declaration section of the application system designated for this purpose.
• In scientific publications: Depending on the journal’s policy, it must be included in the “Author Contribution Statement,” “Acknowledgments,” or a separate “Open Access Declaration” section as specified by the journal. Many international journals now require this declaration.
• In theses and academic papers: The purpose and scope of use must be clearly stated in the preface or methods section.
• In corporate reports: It should be included at the end of the relevant section or as a footnote.

7. Key Risks and Necessary Precautions

The key risks that applicants may encounter, their descriptions, and the preventive measures to be taken against these risks are summarized in the table below:

8. Rules Regarding the Use of AI as an Evaluator and Reviewer

This section applies to all academic staff assuming any evaluator role, including reviewers and panelists evaluating TÜBİTAK applications, members of thesis committees within universities, scholarship evaluation committees, and journal reviewers.

8.1. Fundamental Rule: The Use of ÜYZ in Evaluation Processes Is Strictly Prohibited

TÜBİTAK evaluators are strictly prohibited from using social media platforms for any purpose related to their evaluation duties. This rule applies to all stages of the evaluation and monitoring process, including individual evaluations, panel meetings, reporting, and correspondence. This prohibition is fully consistent with the policies adopted by leading funding agencies such as the NIH, NSF, and the European Research Council (ERC).

8.2. Legal and Ethical Justifications for the Prohibition

Breach of confidentiality: Project proposals contain unpublished original ideas, methodologies, and personal data belonging to applicants. Uploading any part of this content to external ÜYZ tools violates the evaluator’s confidentiality obligation under TÜBİTAK regulations.

Violation of the Personal Data Protection Law (KVKK): The names, contact information, and CVs of researchers included in the application materials constitute personal data under the KVKK. Transferring this data to external platforms subjects the evaluator to legal liability due to the lack of explicit consent and non-compliance with cross-border transfer rules (Article 9).

Risk of FSM violation: Inventions and original ideas in the evaluated project for which no patent application has yet been filed may indirectly become public knowledge by being included in the training data of the ÜYZ model.

Compromise of evaluation quality and integrity: The TÜBİTAK evaluation process relies on the in-depth knowledge, experience, and critical reasoning skills of subject matter experts. AI tools cannot replace this expert judgment; potential algorithmic biases may negatively affect evaluation results.

Accountability: The lack of transparency regarding the extent to which AI influences the creation of the evaluation report undermines accountability. Responsibility for the final evaluation always rests with humans; AI cannot share this responsibility.

Sanctions under the TÜBİTAK AYEK Regulation: Article 9/1-ı of the AYEK Regulation defines acts of negligence or abuse of duty by those involved in evaluation processes as ethical violations. The use of prohibited AI-based evaluation tools is assessed under this article, and if a violation is found, the necessary sanctions are imposed in accordance with AYEK procedures.