Disadvantages of AI-Assisted Code Generation in Software Development
IOSR Journal of Computer Engineering · 2026-03 · 28 (3) · 22-28
DOI: 10.9790/0661-2803022228 ↗Kurz gefasst
Fachartikel in der Zeitschrift IOSR Journal of Computer Engineering, 2026-03. Verfasst von Feruzbek Azimov und 1 weiteren Autor:innen.
Abstract (Originalsprache)
Despite being used by over 80% of software developers, AI coding tools have been quick to integrate into professional software development but slow to be studied. In this article, some of those drawbacks are discussed. According to PRISMA guidelines, 28 primary studies were retrieved from IEEE Xplore, ACM Digital Library, Scopus and Google Scholar between 2020 and 2026, and categorized into six categories. Security: About 40% of the code generated was vulnerable, and 62% under formal verification. Code quality and technical debt: churn more than doubled and refactoring was cut in half. Skill atrophy: Randomized trial showed a 17 point comprehension loss of AI-assisted learners. Privacy and IP risk: unresolved copyleft license propagation. Productivity paradox: Developers were 19% slower, but thought they were 20% faster. Ambiguous responsibility for AI errors at the workplace. Such instruments can bring about local productivity gains, but the literature points to a number of reinforcing risks. Most research has been short-term and lab-based and thus provides only a limited perspective.
Zitieren
APA
Azimov, F., & Senyurek, E. (2026). Disadvantages of AI-Assisted Code Generation in Software Development. IOSR Journal of Computer Engineering, 28(3), 22-28. https://doi.org/10.9790/0661-2803022228
Vancouver
Azimov F, Senyurek E. Disadvantages of AI-Assisted Code Generation in Software Development. IOSR Journal of Computer Engineering. 2026;28(3):22-28. doi:10.9790/0661-2803022228
BibTeX
@article{Azimov2026,
author = {Azimov, Feruzbek and Senyurek, Edip},
title = {Disadvantages of AI-Assisted Code Generation in Software Development},
journal = {IOSR Journal of Computer Engineering},
year = {2026},
volume = {28},
number = {3},
pages = {22--28},
doi = {10.9790/0661-2803022228}
}
EUVIMED