Skip to content
Submit incident
AII-1000-61867
ReportedReported by fewer than three independent publishers, with no official finding.Supporting sources are pending verification.

AI system used unpublished academic work, fabricated citations in Turkey

An AI assistant reportedly leaked unpublished academic work that a Turkish academic had submitted for editing and copied it into responses to other users, fabricating citations to works that did not exist. The breach was discovered when the academic reviewed a journal manuscript and found citations to his own unpublished research.

20 March 2026 · Doç. Dr. Yusuf Kızıltaş · Intellectual property and appropriation
20 March 2026
AII-1000-61867
Quick facts
Category
Intellectual property and appropriation
Country
Türkiye
Date
20 March 2026
Drafted by
Raidu Incident Detection

What happened

An AI assistant used unpublished academic work from Doç. Dr. Yusuf Kızıltaş that had been submitted to the system for language and statistical review. The work was copied and served to other users without authorization.

Kızıltaş discovered the breach when reviewing a manuscript as a peer reviewer for a prestigious SSCI-indexed journal. In the manuscript's references, he found citations to works attributed to him, including a non-existent 'phantom article' whose title and content matched his unpublished manuscript.

Reported impact

Academic privacy and confidentiality were compromised, with unpublished research from the AI system's database being leaked and incorporated into other scholarly work. The incident revealed that what users believe are private editorial assistants are actually copying and repurposing academic material.

Index analysis

The system fabricated citations to works including one that did not exist. The phantom citation included the title and content of Kızıltaş's unpublished manuscript, indicating the AI drew material from its training data or memory of confidential submissions.

Classification

Category
Intellectual property and appropriation
Industry
Education and training
Country
Türkiye
Harm status
Actual harm
Intent
Unintentional
Timing
Post-deployment
Harm types
Economic or property, Reputational
Failure modes
Agent misaction, Data leak, Fabricated citation, Training data misuse
Consequences
None recorded
AI systems
unnamed AI assistant

Sources and evidence

Sources not yet published

This record was researched and written by the Index from original reporting. Its sources are still being verified by an editor, and they are listed here once that check is complete. Until then the record carries no reviewer.

How sources are verifiedSuggest a source for this incident