How Dangerous Can AI Be If Poorly Managed?
Year: 2026
Type: preprint
Abstract: This paper examines the potential harms of poorly managed artificial intelligence (AI), focusing on five major areas of risk: bias and discrimination, deepfake fraud and misinformation, cybersecurity threats, economic and social disruption, and loss of control over advanced AI systems. Using the International AI Safety Report 2026 as a framework, the paper reviews existing research, reports, and documented cases to distinguish between harms already observed in practice and risks that remain uncertain or have primarily been demonstrated in laboratory settings. The discussion highlights how inadequate data quality, insufficient testing, weak safeguards, and limited human oversight can increase the risks associated with AI deployment. It also examines layered approaches to AI risk management, including independent evaluations, technical safeguards, incident reporting, human oversight, public education, and adaptive regulation. The paper concludes that the consequences of AI depend not only on technological capabilities but also on how systems are designed, tested, deployed, and governed. Document type: Literature-based analytical review / preprint.
Author: Ratul, Md Sadman Sakib
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7234324205
