Users' Awareness of Sharing Sensitive Information with AI Tools: A Study among University Students in Mogadishu, Somalia
DOI:
https://doi.org/10.70806/f3zyeg18Keywords:
artificial intelligence, ChatGPT, data privacy, privacy paradox, data protection, SomaliaAbstract
The rapid diffusion of generative artificial intelligence (AI) tools such as ChatGPT has transformed how university students access information, yet it has simultaneously raised pressing concerns about the disclosure of sensitive personal data to conversational AI systems. This study examines the awareness, behaviours, and risk perceptions of university students in Mogadishu, Somalia, regarding the sharing of sensitive information with AI tools. A cross-sectional descriptive survey was conducted between January and April 2026 with 309 respondents recruited through a systematic sampling approach, and data were analysed using Python (pandas, scipy) with descriptive statistics and Pearson's chi-square (χ²) tests of independence. The findings show that ChatGPT was overwhelmingly the most-used tool (89.0%, n = 275), and 50.8% (n = 157) of respondents used AI tools daily. While most information shared was educational (54.4%, n = 168) or general queries (54.4%, n = 168), a notable minority disclosed work-related (13.9%, n = 43) and outright sensitive financial or health data (4.9%, n = 15). Awareness was mixed: respondents were almost evenly split on whether they read AI privacy policies (49.5% Yes), and 29.1% admitted to having shared personal information with an AI. A majority reported low confidence in AI privacy (51.1%, n = 158) and doubted that the government holds sufficient power to protect them (47.2%, n = 146), with mean trust in AI companies of only 5.55 out of 10. Chi-square analysis revealed that reading privacy policies (χ² = 22.68, p < 0.001) and privacy confidence (χ² = 32.37, p < 0.001) varied significantly by education level, as did the disclosure of personal information (χ² = 12.28, p = 0.031). The study evidences a clear privacy paradox: high adoption coupled with limited protective behaviour and low institutional trust. It recommends targeted digital-literacy and data-protection education, and the operationalisation of Somalia's Data Protection Act (Law No. 005 of 2023) and the Data Protection Authority to safeguard citizens in the AI era.