“Trust Me Over My Privacy Policy”: Privacy Discrepancies in Romantic AI Chatbot Apps
Bibliographic record
Abstract
Artificial intelligence (AI) is being pervasively integrated into various facets of human life, including the emotional realm. Romantic AI chatbots, positioned as artificial companions offering emotional support and connection, have witnessed a significant rise in recent years. Users of romantic AI chatbots often reveal personal information during intimate conversations, potentially unaware of the consequences or how their data may be utilized. Complicating matters, lengthy and convoluted privacy policies are commonly overlooked or misunderstood by users. This study aims to address these privacy concerns by introducing a comprehensive framework for analyzing the privacy practices of romantic AI chatbot apps. Through a combination of static and dynamic analysis, we investigate 21 Android romantic AI chatbot apps for: discrepancies between privacy policies and chatbot responses to questions regarding privacy practices; social login and age verification mechanisms; permissions requested by apps; data sharing practices; tracking services employed; and potential security vulnerabilities. Our findings highlight the prevalence of discrepancies between chatbot responses regarding users' privacy and the privacy policies of the apps. Additionally, we note some concerning observations related to: customer service responses to privacy concerns; inadequate age verification measures; contradictions in data sharing claims; and extensive usage of tracking services. We found that all romantic AI chatbot apps tested had discrepancies between their chatbots' responses and privacy policies. None of the apps take any measures against faking the birthdate, and most would continue the conversation despite knowing that the user is underage. 13 out of 21 romantic AI chatbot apps use at least 3 tracking services, and 18 out of 21 apps send detailed device information to tracking services. This study reveals privacy and security flaws in romantic AI chatbot apps, stressing the need for better transparency and user protection measures. Particularly, Discrepancies between chatbot responses and privacy policies highlight the importance of clear communication on data handling.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.013 | 0.068 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.002 | 0.003 |
| Scholarly communication | 0.003 | 0.004 |
| Open science | 0.001 | 0.004 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".